Definition, excuting and evaluating of verification and validation tests (vehicle dynamics) for a vehicle in audex project

For RC cars in the AUDEx project, this means ensuring that the dynamics captured in the model match the observed performance of the RC cars on physical test tracks, such as their respons

Problem Statement

With the rapid advancement of automotive technologies, ensuring the accuracy and reliability of vehicle dynamics is becoming increasingly challenging Key issues include:

Complexity in Testing Environments: Integrating real-world and virtual test environments poses significant challenges in terms of synchronization and data consistency

Data Accuracy and Reliability: Collecting precise data from dynamic systems like vehicles is difficult due to noise and system limitations

Lack of Standardized Frameworks: Existing methods often lack standardized procedures for V&V processes, leading to inconsistent results

Addressing these challenges requires a robust framework capable of handling both virtual simulations and physical tests while ensuring data accuracy and reliability.

Scope and Contributions

The scope of this thesis encompasses the following areas:

• Hardware Integration: Development and calibration of ESP-WROOM-32 microcontrollers and BNO055 sensors for real-time data acquisition

• Software Design: Implementation of data transmission protocols and simulation workflows using CANBUS and the MANEUVER library

• Test Plan Development: Creation of standardized V&V methodologies tailored to the AUDEx project’s objectives

Establishing a comprehensive test framework that integrates hardware and software components

Enhancing data acquisition techniques for improved reliability and accuracy

Providing insights into the effectiveness of V&V processes by analyzing correlations between real-world and virtual test results

Copies for internal use only in Phenikaa University

Source: Sargent (2010) - Verification and Validation of Simulation Models

Figure 2: Simplified version of the modeling process

Fundamentals of Vehicle Dynamics

Definitions of Concepts of Verification and Validation

Verification and validation (V&V) are essential processes for ensuring the reliability and accuracy of vehicle dynamics simulation models In particular, within the AUDEx project involving RC cars, verification confirms that the models are correctly implemented and free of errors These steps are crucial for achieving high-fidelity simulations that can accurately predict real-world vehicle behavior, thereby supporting effective development and testing in automotive research.

5 computational model for vehicle dynamics behaves as expected, solving the system’s equations correctly This might involve ensuring that the model of the

RC car’s motion is mathematically sound, particularly under different operating conditions

A similar scoring approach to increase the credibility of simulation models is developed by NASA [1] The approach categorizes V&V stages and assesses the exerted effort according to a rigor scale for each of the categories This creates an easy to handle overview of the V&V work for the decision makers

Validation involves comparing the model's output with real-world RC car data to ensure accuracy This process includes verifying that simulated car dynamics—such as steering response, acceleration, and handling—align closely with the behavior of the actual RC car during controlled testing Accurate validation is essential for developing reliable and realistic RC car simulations.

Together, these two processes guarantee that the simulation tool can be trusted for future predictions and experiments involving the RC cars in the AUDEx system.

Philosophical Aspect

The philosophical perspective on V&V for vehicle dynamics models touches on the nature of simulation accuracy Since RC cars are scaled-down models of real vehicles, their dynamics should be similar but may involve additional simplifications Verification focuses on the internal correctness of the model, such as ensuring that the forces acting on the car (e.g., tire forces, inertial forces) are correctly calculated based on its design

Validation extends beyond technical correctness to ensure that the model accurately represents real-world phenomena For RC cars in the AUDEx project, this involves verifying that the simulated dynamics align with the actual performance on physical test tracks This includes assessing how well the model predicts the RC cars' responses to steering, braking, and cornering maneuvers, ensuring realistic and reliable simulation outcomes.

Exploring different scientific perspectives reveals distinct approaches: Popperian falsificationism views theories as invalid until falsified, considering them corroborated beforehand; Quinean holism suggests that models exist on a continuum of usefulness and can be revalidated through auxiliary hypotheses or modifications when challenged Meanwhile, Putnam's realistic pragmatist stance emphasizes that simulation models with a history of accurate predictions are valid or approximately valid, supporting a realist interpretation of scientific models.

This work adopts a Popperian perspective, emphasizing that simulation models are inherently approximations of reality and can only be invalidated, not confirmed as fully valid A simulation model cannot be truly valid; it can only be considered "not invalid" if it withstands falsification Therefore, in this context, the term “valid” is defined as “not invalid,” highlighting the provisional nature of simulation model validation under falsifiability principles.

The Popperian falsificationist approach to validating simulation models is vulnerable to two main behaviors: model cooking and unintentional self-deception Model cooking occurs when a simulation model is intentionally adjusted to produce desired or expected results, compromising its integrity Unintentional self-deception, also known as “student’s syndrome” by Carson, involves biased analysis of response data, often focusing on the first signs of apparent validity These problematic practices are common in some validation methods, as discussed in the following section.

Appriaches to Verification and Validation

There are diverse perspectives on validation and verification (V&V), including how they should be conducted and the criteria for considering a model valid Various methodologies and approaches exist, highlighting the importance of selecting appropriate V&V techniques to ensure model accuracy and reliability Understanding these differing viewpoints is essential for establishing effective validation and verification processes in modeling practices.

Carson [1] provided a simple framework for validation of production plant simulation models introducing practical techniques and guidelines, and categorization of modeling errors A “guilty until proven innocent” stance is

7 embraced, the philosophical meaning of which is explored in the following section It is defended that a model can only be deemed valid if it can serve all the purposes it met within the limits of the depth of the dynamics represented within the model That is, if a simulation model of a certain dynamical depth provides satisfactory results in the application it is intended for, but fails to deliver valid results for another process for which the necessary model depth it possesses, the model cannot be valid In the presented framework can be summarized in three consecutive steps as, testing the simulated results for face validity (i.e if they are reasonable), testing the simulation over a range of input parameters, and finally comparing the simulated results to the reference results (Either from an experiment or from a previously validated model) In this comparison step, the results are compared on a reasonable basis if only one data set is available, and a statistical analysis is performed otherwise

This article introduces a comprehensive categorization of modeling errors in simulation projects Modeling errors are primarily classified into project management errors, which stem from faulty planning and process execution; data and data modeling errors, resulting from incorrect data sources, false assumptions about data, or human errors during data entry; logic modeling errors, caused by mistakes in the coding of simulation software; and experimentation errors, arising from the faulty execution of experiments Understanding these categories helps improve accuracy and reliability in simulation modeling.

Using conserved quantities such as momentum, heat, and kinetic energy throughout the system is a key validation approach that ensures model accuracy This method involves analyzing sampled simulation results to verify expected behavior during specific operational modes, helping to confirm whether the model correctly conserves these quantities across the hierarchy Implementing unit testing at the fundamental level allows for effective verification and validation (V&V) by solving simplified, well-known equations to accurately determine the model's response and identify potential issues early in the development process.

Sargent explores four approaches to managing and planning Verification & Validation (V&V) efforts and identifies two paradigms that connect V&V processes to model development The article defines various validation techniques and discusses key aspects such as conceptual model validity, model verification, operational validity, and data validity Additionally, it provides methods for effectively documenting validation results to ensure transparency and reliability in the modeling process.

8 and accreditation is briefly discussed

Various approaches to V&V can be applied to the modeling of vehicle dynamics for the AUDEx project

Model development often involves subjective decision-making by the development team, who base their choices on various tests and results However, this approach can lead to issues such as model cooking and self-deception, potentially compromising the model's accuracy and reliability.

The model relies on the subjective decision-making of the user, who actively participates in the development process to assess its validity User involvement enhances the credibility of the model compared to purely automated methods, though it remains inherently subjective.

Independent V&V (IV&V) involves a third party conducting verification and validation, making it ideal for large projects with multiple development teams This approach can be performed concurrently with development, which may extend project timelines, as subsequent steps often depend on prior validation Alternatively, IV&V can be done after model completion, carrying the risk that earlier errors may cause significant delays Although more costly and time-consuming, IV&V provides higher model credibility and ensures a more reliable validation process.

Scoring in simulation modeling involves assigning subjective scores to various performance aspects, a method widely documented in the literature [5] This approach allows for the evaluation of individual components and the overall model performance by scoring specific categories Alternatively, performance measures can be categorized and scored to derive a comprehensive overall score, facilitating effective assessment and comparison of simulation models.

If the simulation model and its components’ (or each of the categories’) score are above a predetermined pass mark, the model is deemed valid

A potential limitation of this technique is that the model may meet the acceptance score while still harboring overall deficiencies that render it unusable Additionally, the criteria for defining weights and passing scores are highly subjective, which can impact the reliability and accuracy of the evaluation process.

The "A-B-C-D Method" is a comprehensive approach to finite element simulation modeling that emphasizes verification and validation levels through a cost-risk analysis framework It begins with planning (A), followed by solution verification (B), model validation (C), and extrapolated validation beyond the initial scope (D) This method incorporates a scoring system to evaluate validation levels, recognizing that 100% validation is unattainable and that validation depth depends on the application's scope Achieving higher validity entails costs that are carefully analyzed in relation to the specific application needs Additionally, combining multiple verification and validation methods is recommended, as using more than one approach helps address individual weaknesses, thereby enhancing the overall credibility and reliability of the simulation model.

Verification and Validation in Automotive Engineering

Vehicle Dynamics and Modeling

Vehicle dynamics involves the study of how forces and moments affect the motion of a vehicle The modeling of vehicle dynamics in the context of RC cars focuses on creating a representation that can predict behavior in scenarios like turning, braking, or accelerating

The theory of vehicle dynamics is well established, serving as a foundational understanding for the development and validation of simulation models This study focuses on devising and demonstrating a methodology for the validation process rather than creating new simulation models itself Therefore, detailed vehicle modeling theory is not included as a separate chapter; instead, the section highlights various sources related to vehicle dynamics and simulation to support the validation approach.

Vehicle dynamics is a fundamental field within dynamics and control engineering, focusing on the equations that govern the forces and moments acting on vehicle components It examines how these external forces influence the vehicle's response and inertial properties, enabling a better understanding of vehicle behavior The overall motion of vehicles is primarily driven by factors such as steering inputs, acceleration, braking, and external environmental conditions Studying vehicle dynamics is essential for enhancing vehicle safety, stability, and performance across various driving scenarios.

The tire-road contact surface experiences ten horizontal and vertical forces resulting from driver inputs and road conditions, which are transmitted to the vehicle body through suspension, elastic bushings, and steering systems These interconnected components form a complex, high-order nonlinear system characterized by intricate geometric relationships, force elements, and viscoelastic behavior, with many parameters that are difficult to measure directly Vehicle dynamics primarily focus on longitudinal performance, lateral stability, handling, and vertical ride comfort While longitudinal and lateral forces can be decoupled in practice, vertical forces typically interact with other motion components, such as roll during lateral maneuvers and pitch during longitudinal movements, highlighting the complex interplay in vehicle behavior.

Vehicle dynamics simulation plays a crucial role in the automotive industry, supporting the development and modification of vehicle models, research in simulators and ergonomics, and the advancement of mechatronic components The fundamental single track (bicycle) model, introduced in 1940 and effective up to 0.4 g lateral acceleration, remains widely used today Key milestones include Olley's 1946 work on vehicle characteristics like understeer and Segel's 1956 model for time domain analysis, with frequency domain response later explored by McRuer et al in the 1970s Extensive literature exists on vehicle dynamics fundamentals and applications, covering tire and brake systems, engine management, and modeling techniques, with Lugner and Plühl providing a comprehensive overview of simulation methods and model types.

One of the most important sources on tire dynamics is written by Pacejka, [13] who also developed the so called Magic Formula, an empirical tire model which relies on curve fitting using experimentally measured tire data, which is also the tire model used in this work Further work on tire dynamics and tire modeling

Numerous tire modeling approaches exist, varying in complexity and application scope Analytical models, like Rill’s first-order tire model based on Taylor expansion of differential equations and Lacombe’s mechanically inspired model, provide simplified yet effective representations Modal parameter-based analytical models also offer valuable insights Physical models utilizing Finite Element Method (FEM), such as FTire, deliver high accuracy but require substantial computational resources, making them less suitable for real-time applications Tire wear modeling is essential, as wear significantly impacts experimental accuracy and is crucial for applications like race performance prediction, tire development, and fleet management.

For the AUDEx project, this involves developing both the physical models (e.g., kinematic models that describe the motion of the car) and the force models (e.g., tire models that predict the friction forces) The challenge is to find the right balance between simplicity for computational efficiency and complexity for accuracy.

Practice of Validation of Simulation Models for Vehicle Dynamics

Many publications claiming to validate methodologies often focus solely on applying techniques to specific cases, resulting in project-specific validation in vehicle dynamics This section explores such sources along with other relevant research that discusses validation methods without presenting comprehensive validation studies.

This methodology involves three key phases: collecting experimental field data, measuring independent vehicle parameters, and developing the model formulation Emphasizing the importance of independent parameter measurement, the model parameters should remain unchanged during field tests to ensure integrity Validation is achieved by comparing simulation predictions with field data using the same driver control inputs These comparisons are conducted in both time and frequency domains; the time domain analysis assesses steady-state and low-frequency responses, including nonlinear effects, while the frequency domain evaluation focuses on high-frequency dynamics during transient conditions.

In another study by McNaull et al., [17] a heavy truck simulation model was first modified according to comparison of experimental and simulation results for lateral steady state maneuvers; and then validated for dynamic response using a transient maneuver The work does not introduce or explain the methodology but rather is a demonstration that the end result of the project is successful Visual graphical comparison technique is used for validation, but instead of overlaying the graphs, side-by-side placed diagrams are used, which diminishes the credibility of the validation judgement Also, no metics or statistical analysis are performed The study demonstrates the correct way of using experimental data to correct the simulation model, by determining the steady state offset and then testing the modified system with a transient maneuver On the other hand applied validation technique, side-by-side representation of quantities of interests, somewhat lowers the possibility of a healthy call for validity

Allen et al (2002) emphasize the importance of accurate parameter measurements within the targeted operating regime to ensure the validity of full vehicle models They highlight that for simulations involving limit handling scenarios, such as rollover or tire saturation, subsystems like tires must be characterized under large slip conditions and higher-than-normal loads Additionally, capturing non-linearities caused by larger deflections during highly dynamical maneuvers is crucial for realistic and reliable vehicle simulations.

Effective validation of models requires addressing key issues such as proper formulation, rigorous verification of computer coding, accurate parameter estimation, and reliable measurement procedures Comparing experimental data with simulation results is essential for ensuring model accuracy A comprehensive validation process should encompass both steady-state and transient maneuvers, evaluating outcomes in both time and frequency domains to ensure robust performance across various operating conditions.

The maneuvers are so sequenced; first quasi steady state, then step response, then pulse response (evaluated in frequency domain) and finally a purpose

In this study, 13 real-world maneuvers, such as lane changes, were performed to evaluate the simulation model's accuracy While sine sweep maneuvers could serve as an alternative to pulse response tests, the current validation approach relies on confidence intervals without establishing clear validation criteria or quantitative validation metrics Consequently, validation judgments are primarily based on subjective assessments, lacking objective measures to determine the “adequacy” of the simulation model, which highlights the need for more rigorous and quantifiable validation methods.

This article discusses a valid approach to the validation process, highlighting the importance of accurate parameter estimation and data validity It emphasizes the proper sequencing of test maneuvers—from steady-state to transient and real-life scenarios—to ensure comprehensive testing While sine sweep is suggested as an alternative for frequency response testing due to its consistent power across the frequency range, the current methodology lacks transient response testing in the time domain and does not establish quantitative validation criteria Validation is primarily based on subjective visual assessments comparing test and simulation results.

In the practice of V&V for RC cars, the goal is to ensure that the simulated behavior matches the real-world performance For example, when simulating the

RC car’s response to a steering input, validation involves comparing the model’s output with real-world tests where the RC car is driven through the same maneuver

The main practices for validating the simulation models of RC cars include:

Field testing RC cars involves running them on designated test tracks and comparing the actual performance results with simulation predictions Key tests include acceleration runs, cornering assessments, and maneuverability evaluations, providing valuable insights into the vehicle's real-world capabilities and helping to refine design and performance.

• Sensor Data Integration: For more sophisticated validation, integrating sensors (such as accelerometers, gyros) into the RC cars can provide real-time data that can be used to validate the simulation model’s

14 predictions These sensors capture real-world forces and vehicle motion that can be compared to the simulated results

Model calibration is a crucial process where simulation model parameters are adjusted using experimental data to enhance predictive accuracy For instance, refining tire parameters in the simulation based on observed slip angles from real-world tests ensures more reliable and precise simulations This optimization improves the model's ability to accurately predict real-world behavior, making calibration essential for robust simulation studies.

Theory of Validation of Simulation Models for Vehicle Dynamics

Research on methodologies for validating vehicle dynamics is limited and lacks diversity Hoskins and El-Gindhy [19] conducted a comprehensive literature review that highlights existing validation approaches for vehicle dynamics models, particularly those used in driving simulators.

The 1990 paper by Heydinger et al is a foundational work in the field, as it is arguably the first study to introduce a comprehensive validation methodology for vehicle dynamics simulation models This pioneering research has significantly contributed to establishing standardized approaches for assessing the accuracy and reliability of vehicle simulation tools Its insights continue to influence the development and validation practices within vehicle dynamics engineering, making it a crucial reference for researchers and practitioners alike.

According to this reference, validation is defined as showing that, within some specified operating range of the vehicle, a simulation’s predictions of a vehicle’s responses agree with the actual measured vehicle’s responses to within some specified level of accuracy This definition emphasizes three points:

• A simulation’s predictions may only be correct within some portion of the system’s operating range (e.g a lateral acceleration range, or a steering angle input frequency interval)

• A simulation’s validity is determined for a specified group of inputs and outputs (e.g a validated lateral dynamics model with suspension degree of freedom is not necessarily valid for comfort studies)

• A simulation’s validity is determined according to the variance between the simulation’s outputs and experimental measurements

The described method uses repeated experimental runs at each test condition to

To ensure accurate statistical analysis and reliable confidence intervals, generate sufficient data in both time and frequency domains to account for random errors in experiments Employ qualitative and quantitative methods to compare simulation predictions with actual test measurements, using visual graphical comparison for clear and effective validation of results.

Accurate center of gravity (CG) height measurement is crucial for vehicle dynamics validation, as errors—even if within confidence intervals for yaw rate and lateral acceleration—can be detected through lateral load transfer analysis For maneuvers like the fishhook, where lateral load transfer significantly impacts vehicle behavior, CG height, roll angle, and lateral load transfer should be established as key validation metrics from the outset It is essential to incorporate these parameters during the planning phase of the simulation, ensuring the target maneuver is thoroughly analyzed and the appropriate test maneuvers and validation metrics are selected, rather than addressing these issues only in the final validation stage.

Inaccurate road friction coefficient values used in simulations can lead to unreliable results, as these values often do not match those obtained during real-world testing Relying solely on manufacturer-supplied data or calibrating based on test vehicle data may mask underlying errors, underscoring the importance of using precise, field-verified friction coefficients for more accurate simulation outcomes.

The theoretical aspects of validation for vehicle dynamics focus on creating a robust framework for comparing simulated and real-world data In the context of

RC cars, the theory emphasizes the importance of understanding the uncertainties involved in both the model and the experimental measurements

• Uncertainty Quantification: Acknowledging that all models contain some degree of error or uncertainty, whether in the model itself or in the experimental data used for validation

Error metrics like mean squared error and correlation coefficients are essential for assessing how accurately a model represents real-world data These metrics provide quantitative measures of model performance, enabling researchers to evaluate its accuracy and determine if further refinement is necessary Utilizing appropriate error metrics ensures robust model validation and improves predictive reliability, guiding iterative improvements in modeling efforts.

• Sensitivity Analysis: Understanding how sensitive the model’s output is to changes in the input parameters In the case of RC cars, this could involve testing how changes in suspension parameters or tire properties affect the simulated and real-world behavior

Vehicles typically exhibit zero degrees of roll deflection under trim conditions If a simulation of straight driving at constant speed on a frictionless surface shows a non-zero roll angle, it suggests an error in at least one parameter influencing static roll deflection, assuming the mathematical model is correct Common causes include inaccuracies in tire geometry or stiffness parameters, or incorrect spring rate values.

On the other hand, what if parameter data of one of spring rates and one of the tires are mistakenly entered at the same time, in such a way that their effects at trim condition cancel out each other? In this case, a more dynamical maneuver (e.g steady state cornering), individual load or force measurements for the tires, or an isolated test case would be more practical In the first two of these suggestions, there is absolutely no guarantee that the simultaneously wrongly entered (or measured) parameters can be identified Concerning the third suggestion, generally speaking, it is impossible to devise a maneuver which would isolate every parameter of the system since most of the parameters are inherently interacting One can only come up with a limited number of such maneuvers (for example lateral and longitudinal maneuvers can be separated, but the vertical dynamics almost always affect the other two) but as previously said, there is no guarantee such an error can be detected Nevertheless this approach is very useful in increasing the model confidence

Allen et al provided [21] a methodical approach to the validation problem Possible problem areas causing inconsistencies between computer models and real world are described as:

It is advocated that the vehicle dynamics model validation must be considered in context and defined in terms of the domain of useful application, since a simulation model can only be valid up to a degree and a model should be aimed for a certain behavior, and a valid model according to analysis of general system response does not guarantee valid subsystems models

Validation method presented in this work is summarized in four steps:

• Conceptual validity of the mathematical model

• Face validity (reasonableness) of the simulation model response

• Consistency of input, intermediate and output variables

• Agreement between the simulated behavior and the reference system (real or simulated)

The article evaluates lateral direction validation using three test cases: steady-state cornering, sinusoidal sweep, and lane change maneuver However, it lacks a clear methodology for assessing the results, as the study does not define validation metrics, apply statistical methods, or establish validity criteria, which limits the reliability of the findings.

Vehicle dynamics simulation models can be validated through three primary approaches Heydinger et al [22] and Garrott et al [23] emphasize operational validity by comparing test measurements with simulation results to ensure real-world accuracy Bernard et al [24] and Gruening et al [25] recommend using analytical solutions and face validity checks primarily for parameter identification and error detection Additionally, Allen et al highlight the importance of face validity, analytical solutions, and comprehensive validation techniques to enhance model reliability and predictive power Incorporating these approaches ensures robust validation of vehicle dynamics models for accurate performance prediction.

18 common sense checks with less methodical approach to vehicle testing.

Expert Views on the Subject

In the field of vehicle dynamics, experts emphasize that validation is not a one- time event but an ongoing process, especially for systems like RC cars in the AUDEx project Some expert views include:

Real-world validation is essential for accurate RC car performance assessment, as experts point out that high-fidelity simulations, while helpful, cannot fully replace actual testing Factors such as battery power, road surface, and environmental conditions significantly influence RC car behavior, emphasizing the importance of ongoing field testing Continuous validation ensures reliable results and optimal performance in real-world scenarios.

• The Role of Simulation in Optimization: While validation is necessary, experts also highlight the importance of using simulation models as a tool for optimization In your AUDEx project, simulation allows for testing a wide variety of configurations and scenarios before physical testing, helping to refine the design of both the RC cars and the simulation model

Collaborative validation and verification (V&V) efforts are essential for enhancing the accuracy of vehicle dynamics simulations Experts emphasize the importance of sharing data and findings across different projects and institutions to develop standardized simulation methodologies This collaborative approach not only improves the quality of simulation results but also fosters the advancement of industry-wide standards, ensuring more reliable and precise vehicle dynamics analysis.

Tools and Frameworks: CANBUS, MANEUVER, and Testing Platforms

CANBUS

The Controller Area Network (CAN) bus is an essential communication protocol widely used in automotive engineering for transmitting data between various electronic control units (ECUs) within a vehicle In the context of the AUDEx

The 19 project utilizes RC cars equipped with CANBUS technology to enable seamless communication between multiple vehicle subsystems, including sensors, motor controllers, and actuators CANBUS ensures high reliability and robustness, which are essential for real-time data transfer in dynamic environments, enhancing the vehicle's performance and safety.

CANBUS operates at high speeds of up to 1 Mbps, enabling real-time transmission of vehicle data such as speed, acceleration, steering angle, and sensor readings In RC cars used in this project, CANBUS facilitates seamless data transfer between components like the BNO055 sensor and motor controller, supporting continuous monitoring and precise control This real-time data capability is essential for validating vehicle dynamics models, as it allows for accurate data collection during test maneuvers to compare with simulation results.

MANUEVER

MANEUVER is a powerful vehicle dynamics modeling and simulation tool that enables detailed analysis of vehicle behavior in various driving scenarios Its advanced framework simulates dynamic responses during cornering, braking, and acceleration, allowing engineers to evaluate vehicle performance under diverse conditions With MANEUVER, users can explore real-time vehicle responses to different driving inputs, making it an essential tool for optimizing vehicle design and safety.

To validate the simulation model replicating a vehicle's response in complex, real-world maneuvers, it is essential to simplify the test process This is achieved by dividing intricate maneuvers into standardized, repeatable test cases that improve comparability and realism through "pseudo-driving" scenarios Employing these simplified maneuvers ensures accurate assessment of the model’s performance under controlled, repeatable conditions, aligning with best engineering practices.

The standardized maneuvers are the maneuvers which clearly exhibit general

20 dynamic characteristics of a vehicle, and are economically feasible, experimentally repeatable and consequentially comparable, but nonetheless in real life nonexistent They are targeted in simulation studies because of the aforementioned qualities they possess, in order to increase the confidence in a simulation model that it is not invalid, so that it can be used in pursuance of predicting the response of a vehicle to a by no means experimentally repeatable real world maneuver which is turbulent due to the previously mentioned factors This relationship between the standardized test maneuvers (STMs), simulation models and real world is depicted in this Figure which demonstrates the so- called validation triangle Here, the real event is deconstructed and simplified to a standardized test maneuver, the results of which are used in the validity assessment of the simulation model, which aims to predict the outcome of the real event

The standardized maneuvers can be classified in different ways: According to their relevant analysis domains, input methods, and their scope of application

A key method for classifying lateral dynamics test maneuvers is based on the way the input is introduced This classification depends not on whether a test driver or a driving robot performs the maneuver, but on how the input is defined If the input is specified independently of any system or environmental factors, the maneuver is categorized as an open-loop maneuver.

Open loop test maneuvers encompass a variety of techniques such as steady-state cornering, braking in a turn, and power-off reactions in a turn, as cited in references [26], [27], and [28] Additionally, they include step response maneuvers, single sine input, sine sweep methods, and sine with dwell maneuvers, as outlined in references [29] and [30] During these tests, the steering input is predefined and remains constant throughout the experiment to ensure consistency and accurate data collection.

The AUDEx project relies on MANEUVER as a vital tool for simulating RC car performance, enabling detailed analysis of handling characteristics such as roll, pitch, yaw, and stability by inputting parameters like vehicle mass, tire properties, and suspension settings This simulation capability helps evaluate RC car behavior during specific maneuvers, including sharp turns and high-speed braking Additionally, MANEUVER can model different environmental conditions, like various road surfaces and weather effects, offering a comprehensive understanding of the vehicle’s performance across diverse scenarios.

In conjunction with real-world testing, MANEUVER helps refine simulation models by identifying discrepancies between simulated and observed behaviors, which are then addressed through model adjustments and recalibrations.

Testing Platforms

Testing platforms are essential for validating vehicle dynamics models, offering environments for physical testing and data collection For the AUDEx project with RC cars, these platforms encompass both physical and virtual settings, each playing a vital role in model validation.

Physical testing is essential for gathering real-world data by running RC cars on controlled tracks, capturing key vehicle motion parameters such as velocity, steering angle, and acceleration through onboard sensors like accelerometers, gyros, and GPS systems This data validates and enhances the accuracy of simulation models created with tools like MANEUVER, ensuring reliable performance predictions Incorporating physical testing into the development process improves the overall robustness and realism of autonomous RC car systems.

Physical testing offers valuable real-world data that can be compared with simulated results, ensuring the accuracy of the RC car's performance assessment It allows for the evaluation of the RC car under actual road conditions, including variables such as road texture, incline, and external disturbances like wind, which are crucial for comprehensive performance analysis.

Virtual testing enhances physical testing by simulating RC car dynamics in a virtual environment, making it ideal for evaluating complex or risky scenarios Using advanced simulation tools like MANEUVER, engineers can accurately model the impact of various tire types, suspension setups, and road conditions on vehicle performance This approach allows for efficient assessment of different configurations, reducing costs and safety concerns associated with real-world testing.

RC car's handling and stability under extreme or hypothetical conditions, such as wet or icy roads, without the need for physical trials

Together, physical and virtual testing provide a comprehensive approach to validating vehicle dynamics models, allowing for the refinement of both simulation tools and real-world vehicle performance.

Related Work and Case Studies

Related Work in Vehicle Dynamics Simulation for RC Cars

Numerous studies showcase the effectiveness of simulation and modeling tools in RC car dynamics, emphasizing their role in autonomous navigation, performance optimization, and control system development These research efforts highlight the critical need to combine analytical models with real-world data to enhance the accuracy and reliability of RC car performance and control systems Incorporating advanced simulation techniques allows for precise prediction of vehicle behavior, leading to improved design and safer, more efficient RC car operations.

Smith et al (2019) developed a hybrid modeling approach that combines real-time sensor data with physics-based vehicle dynamics models This integration facilitates adaptive control systems, enhancing vehicle performance and safety Such innovative modeling techniques are crucial for advancing intelligent transportation technologies and improving real-time vehicle management.

A 23-system for RC cars can dynamically adjust parameters in response to changing environmental conditions, enhancing performance Jensen and Wang (2018) highlighted the significant impact of tire characteristics, especially friction coefficients, on RC car handling, stability, and maneuverability Their research demonstrated that optimizing tire parameters through simulation tools is essential for improving overall vehicle performance.

Recent studies highlight the critical importance of combining simulation and experimental validation to develop precise vehicle dynamics models For the AUDEx project, utilizing real-time sensor data for calibration and validation is essential to ensure high-accuracy predictions, ultimately enhancing model reliability and performance.

Case Studies in Verification and Validation of Vehicle Dynamics Models 23

Numerous case studies have examined the verification and validation of vehicle dynamics models in both real-world and simulation environments For example, the Ford Simulation and Testing Program (Ford, 2015) validated its vehicle dynamics models by comparing simulated handling test results with actual prototype vehicle data This approach proved that high-fidelity simulations can accurately predict vehicle behavior, while also emphasizing the critical role of environmental factors like road surface roughness and tire wear that can affect model accuracy.

Research by Zhou et al (2021) highlights the significance of real-time data acquisition using CANBUS for analyzing RC car dynamics during controlled test track maneuvers Their study emphasizes the critical role of sensor data integration in improving simulation accuracy, reducing discrepancies between predicted and actual vehicle behavior, and enhancing the reliability of RC car performance models.

Application of Testing Frameworks in Vehicle Dynamics

Testing frameworks like MANEUVER are extensively utilized in vehicle dynamics research for simulating and validating vehicle behavior across diverse driving scenarios The Toyota Research Institute (Toyota, 2019) employed similar simulation tools to model autonomous vehicle dynamics, conducting thousands of virtual tests to assess the vehicle's performance in complex situations such as sudden lane changes and evasive maneuvers These simulations enable researchers to identify and address potential handling issues proactively, reducing the need for costly physical testing and enhancing overall vehicle safety and reliability.

Brown et al (2022) utilized MANEUVER-like testing frameworks to optimize RC car performance for competitive robotics by simulating diverse driving conditions Their approach enabled precise fine-tuning of control algorithms, leading to enhanced vehicle handling This research demonstrates the importance of integrating simulation and physical testing to ensure RC cars operate reliably in real-world scenarios.

These case studies highlight the importance of simulation tools and testing frameworks in validating vehicle dynamics models For the AUDEx project, simulations offer valuable insights for optimizing RC car performance, while real-world testing helps refine and improve these models for greater accuracy and reliability.

Overview of the AUDEx System

The AUDEx project (Automotive Development in 1:x) allows students to perform modern development tasks by working with realistic remote-control vehicles equipped with cameras This initiative provides hands-on experience in applying industry-standard tools and techniques, preparing students for real-world automotive engineering challenges.

Autonomous vehicles integrate advanced 25 microcontrollers, sensors, and actuators to develop and implement complex control algorithms, ensuring precise operation These vehicles can also be controlled through a sophisticated motion system, allowing researchers to assess how autonomous development influences the driving experience for both drivers and passengers.

AUDEx utilizes scaled vehicles to offer a comprehensive, real-world experience of vehicle development, including component and function testing, chassis setup evaluation, and autonomous driving tasks This innovative approach enables seamless integration of virtual simulations and real-world testing, enhancing the development process for automotive engineers.

The AUDEx system consists of the following key components:

• Simulation Framework: Tools such as MANEUVER are used to simulate vehicle dynamics based on input parameters, including vehicle weight, tire properties, and environmental conditions

• Data Acquisition System: This includes sensors (e.g., BNO055), microcontrollers (e.g., ESP32), and onboard data storage (e.g., SD cards) that record real-time vehicle performance data during testing

• Vehicle Models: Both RC cars and real vehicles are used in tandem to validate simulation results RC cars are tested in a controlled environment, while real vehicles provide additional data points for comparison

• CANBUS Communication: The data exchange between the vehicle components, such as sensors and controllers, occurs through CANBUS, which is responsible for real-time data transmission between various subsystems

The AUDEx system's framework supports the iterative process of simulation, real- world testing, and data comparison, allowing for continuous refinement of vehicle dynamics models.

Workflow and Approach

The workflow of the AUDEx project follows a systematic approach to validate vehicle dynamics models through real-world data and simulations The process is outlined in the following steps:

The first step in developing accurate RC car models is data collection, which involves gathering real-world data using onboard sensors like the BNO055 for orientation and acceleration This process is conducted during controlled driving scenarios to precisely test specific vehicle dynamic behaviors, including acceleration, braking, and cornering, ensuring comprehensive data for performance analysis and optimization.

Simulation involves inputting collected data into a virtual environment, such as MANEUVER, to model the vehicle's behavior under similar conditions This process enables testing complex scenarios that are challenging to replicate physically and allows for pre-testing maneuvers safely before executing them on the actual vehicle.

• Data Comparison and Analysis: After running the simulation, the results are compared with the real-world data collected in the previous step Any discrepancies between the simulated results and actual data are analyzed to determine potential areas for improvement in the vehicle dynamics models

Model refinement involves adjusting simulation models to enhance their accuracy, driven by insights from real-world testing Key factors such as tire friction, vehicle mass, and sensor calibration are analyzed and incorporated into the refinement process These updates ensure that the simulation more precisely reflects real-world conditions, leading to more reliable and effective testing outcomes.

Iterative testing involves repeatedly testing, comparing, analyzing, and refining simulation models to ensure accurate prediction of vehicle behavior across diverse driving conditions This continuous feedback loop enhances both virtual and real-world testing by progressively refining the models, resulting in more reliable and precise vehicle simulations.

The final evaluation of vehicle dynamics models involves comprehensive testing to assess their suitability for real-world applications After refining the models through multiple iterations, this stage compares their predictive accuracy against actual vehicle performance data This iterative process ensures high precision, as each stage includes a thorough comparison of simulated results with real-world data, ultimately confirming the models' reliability for practical use.

Hardware and Software Integration

Vehicle Models (RC Cars)

The AUDEx project utilizes two types of vehicle models: RC cars

The use of model allows for extensive testing and comparison under various conditions

Figure 4: RC car with the circuit board

RC cars serve as the primary platform for data collection in vehicle dynamics testing Equipped with advanced sensors such as the BNO055, they precisely capture orientation and acceleration data, enabling detailed analysis Their small size and high maneuverability make RC cars ideal for creating controlled testing environments, which enhance the accuracy and repeatability of vehicle behavior assessments This reliable data collection is essential for validating and refining vehicle dynamics models.

With vehicle types are integral to the project’s approach, allowing for cross- validation of simulation models with real-world data.

ESP32, BNO055 Sensor and SD Card Setup

The hardware setup for data acquisition in the AUDEx project includes several key components that work together to collect, process, and store real-time vehicle performance data

The ESP32 microcontroller is a versatile and powerful component that interfaces seamlessly with the BNO055 sensor and SD card module It efficiently collects data from the sensor, processes the sensor readings, and stores the valuable data on the SD card for further analysis Its capabilities make it ideal for projects requiring real-time data acquisition and storage, ensuring reliable performance and easy integration.

Its Wi-Fi and Bluetooth capabilities also allow for wireless communication, enabling remote monitoring of vehicle performance during testing [32]

The BNO055 sensor is a 9-axis device that measures orientation, including roll (rotation about the vehicle's longitudinal axis), pitch (rotation about the transverse axis), and yaw (rotation about the vertical axis) It also captures linear motion through acceleration data (Ax, Ay, Az) in meters per second squared and rotational motion via angular velocity (Gx, Gy, Gz) in degrees per second or radians per second This sensor is vital for real-time tracking of a vehicle’s movement, offering essential data on stability, directional changes, and acceleration during various maneuvers.

The SD card module is essential for storing collected data during testing, allowing for the recording of large data volumes It logs structured data in accessible file formats, facilitating detailed post-test analysis This enables comparison between real-world performance and simulations, helping to identify discrepancies and refine models for improved accuracy.

Source: Vehicle Dynamics – Martin Meywerk (Chapter 1)

This article discusses two nominal axis orientations recognized in technical terminology: the SAE J670 orientation, adopted from aeronautical conventions, and the ISO 8855:1991 orientation, known as the Z-Up system Both axis systems are equally valid, with the choice depending on the specific requirements of the analysis or testing performed The ISO 8855 axis orientation, or Z-Up, contrasts with the traditional SAE J670 Z-Down orientation, and these different configurations are illustrated in the accompanying figure.

When definitions depend on axis orientation, dual definitions are provided, with the Z-Up axis orientation-based definition listed first Many terms in this section are also referenced and defined in SAE J2047, ensuring clarity and consistency across standards.

Additional frame systems are essential for accurately describing vehicle motion, such as the frame fixed to the wheel's center of mass (𝑺 𝒄𝒎𝒘 , 𝒆⃗⃗⃗⃗⃗⃗ , 𝒆 𝒘𝒙 ⃗⃗⃗⃗⃗⃗ , 𝒆 𝒘𝒚 ⃗⃗⃗⃗⃗⃗ ) and another at the contact patch (𝑺 𝒄𝒑 , 𝒆⃗⃗⃗⃗⃗ , 𝒕𝒙 𝒆⃗⃗⃗⃗ , 𝒕𝒚 𝒆⃗⃗⃗⃗ ) The diagram includes the 𝒕𝒛 frame and angles ψ, ϑ, and ϕ, which represent single rotations rather than a sequence, simplifying the analysis This approach allows for the consideration of individual rotations without accounting for their interactions, though analyzing combined rotational interactions would significantly increase equation complexity.

During straight-line vehicle motion, the vehicle's acceleration and velocity vectors align, focusing on longitudinal dynamics such as resistances, driving performance, braking, and acceleration This movement involves rotational motion about the vehicle's lateral axis, known as pitch, which is directly linked to forward motion and the vehicle's center of mass above the road When accelerating or braking, inertia forces act on the center of mass, causing a pitch moment Additionally, vehicle vibrations caused by uneven roads include translations in the vertical (z) direction, yaw rotations around the vertical axis, pitch rotations about the lateral axis, and roll rotations about the longitudinal axis, impacting ride comfort and vehicle stability.

During cornering, the vehicle experiences a non-constant yaw angle, causing the vehicle's direction vector not to align with the vehicle's longitudinal axis In addition to rotating about the lateral axis (roll), the vehicle also rotates about the longitudinal axis during acceleration or deceleration, highlighting the complex three-dimensional dynamics involved in vehicle handling and stability.

The vehicle's pitch motion occurs along the ⃗⃗⃗⃗⃗⃗ -axis, accompanied by lateral movement during cornering The investigation of cornering dynamics, covered in the third part of this book, highlights the importance of lateral dynamics in vehicle behavior These considerations demonstrate that vehicle motion typically involves multiple degrees of freedom, emphasizing the complexity of analyzing and controlling vehicle stability during maneuvers.

Together, these components form the core of the data acquisition system, enabling accurate and reliable data collection throughout the testing process

Hardware Design and Setup

Sensor calibration and Mounting

The BNO055 sensor is the core component for collecting precise orientation and acceleration data, offering a comprehensive 9-axis measurement that includes gyroscope, accelerometer, and magnetometer readings Proper calibration of this sensor is crucial to achieve accurate and reliable measurements during vehicle maneuvers, ensuring optimal performance of the navigation and control systems.

To successfully install and integrate the ESP32 microcontroller with the BNO055 orientation sensor and SD card module, begin by ensuring proper hardware connections following the manufacturer's pinout diagrams Next, install necessary libraries such as Adafruit's BNO055 and SD libraries in your development environment to facilitate seamless communication Then, write and upload code that initializes the sensor and SD card, verifying each component's functionality through serial debugging Properly configuring the I2C or SPI interfaces and ensuring correct power supply to all modules is crucial for stable operation Finally, implement data acquisition routines to accurately read orientation data from the BNO055 sensor and log this information onto the SD card, enabling efficient data storage and retrieval Following these steps carefully ensures reliable system performance and robust data logging capabilities.

Begin by supplying the required power to all components:

• Connect the 3.3V pin from the ESP32 to both the VIN of the BNO055 and the VCC of the SD card module

• Ensure a common ground (GND) connection between all devices for electrical consistency

Establish the I2C communication lines for the BNO055:

• Connect SDA (data line) of the BNO055 to GPIO21 of the ESP32

• Connect SCL (clock line) of the BNO055 to GPIO22 of the ESP32

These pins are commonly used defaults on the ESP32 for I2C and ensure reliable communication with the sensor

SPI Communication – SD Card Module

Set up the SPI interface for the SD card module using the following connections:

• MISO (Master In Slave Out) → GPIO19

• MOSI (Master Out Slave In) → GPIO23

These connections enable the ESP32 to write sensor data to the SD card efficiently

Before powering on, verify that:

• No GPIO pin is used for multiple connections

• Pins reserved for ESP32 internal flash (such as GPIO6 to GPIO11) are avoided, as using them may lead to instability or failure to boot

Power on your setup and conduct the following checks:

• Measure voltages across modules to confirm correct power delivery

• Use a multimeter to test for continuity and proper wiring connections

This step ensures the circuit is safely connected and ready for software deployment

Finally, install the necessary software libraries and upload test code:

• Install libraries such as Adafruit BNO055, SPI, and SD via your Arduino or PlatformIO environment

• Upload a basic script to verify communication with both the sensor and the SD card

This confirms that the setup is working properly and ready for data acquisition or further development

Figure 6: The circuit board includes: SD card, BNO 055 and ESP - 32

The following calibration steps are undertaken:

Proper sensor alignment is essential for accurate data collection; the sensor must be securely mounted on the vehicle in a position that captures all necessary movement data, including roll, pitch, yaw, and acceleration For optimal performance, the sensor is typically placed at the vehicle’s center of gravity to minimize measurement distortion caused by vehicle movement Ensuring precise sensor placement enhances data accuracy and reliability in vehicle dynamics analysis.

• Calibration Procedure: The BNO055 sensor requires initial calibration before it can be used for data collection Calibration is done using the sensor’s built-

34 in software library to adjust for biases in the sensor readings This involves performing rotation and movement tests to ensure that the sensor’s accelerometer and gyroscope axes align properly

Dynamic calibration is conducted during real-time operation to correct small inaccuracies that may develop during testing This process is essential for maintaining the precision of sensor data, particularly during high-speed or complex maneuvers, ensuring reliable and accurate measurements in real-world conditions.

Sensor testing involves conducting a series of test runs after calibration to ensure sensor readings are consistent and reliable This process verifies that critical data such as roll, pitch, yaw, and acceleration are accurately recorded during actual testing, ultimately improving the precision and dependability of sensor measurements.

After correctly calibrating and mounting the sensor, the system is prepared for data collection The sensor provides orientation and acceleration data, which are processed and logged by the ESP32 microcontroller, ensuring accurate and efficient data management.

Data Logging System: ESP32 and SD Card

The ESP32 microcontroller is used as the central data logger for the AUDEx system It collects sensor data from the BNO055, processes the information, and stores it on an

The ESP32 is an ideal choice for data logging systems thanks to its robust processing power, versatile capabilities, and extensive connectivity options An SD card is used for storing data for later analysis, ensuring that collected information is securely saved and easily accessible Key elements of this data logging system include the efficient integration of SD card storage and the powerful features of the ESP32, making it a reliable and flexible solution for various applications.

The ESP32 seamlessly interfaces with the BNO055 sensor to collect real-time orientation and acceleration data, ensuring precise measurement Each data reading is timestamped, enabling accurate synchronization during analysis and enhancing the reliability of sensor data integration.

The ESP32 efficiently stores collected data on an SD card in a structured, easily accessible format, typically using comma-separated values (.txt) This data includes precise timestamps alongside critical sensor measurements such as roll, pitch, yaw, and acceleration, facilitating seamless post-test analysis and ensuring reliable data management.

• Data Transmission: The ESP32 also supports wireless communication through Wi-Fi and Bluetooth, enabling remote data transmission if needed

For instance, data can be transmitted in real-time to a computer or mobile device for live monitoring

Effective power management is essential for the long-term operation of the ESP32 and sensor systems in vehicle testing By implementing low-power modes and optimizing coding strategies, we significantly reduce energy consumption, ensuring extended battery life during prolonged testing sessions.

After completing the test, remove the SD card to easily retrieve stored data This data can then be imported into analysis software for detailed comparison with simulated vehicle dynamics, ensuring accurate performance assessment and insightful analysis.

The integration of the ESP32 with the BNO055 sensor and SD card creates a flexible and reliable data acquisition system that supports the testing process across various maneuvers.

Software Architecture

Data Transmission and Acquisition

Data transmission and acquisition are vital to ensuring accurate and synchronized data collection during vehicle tests The data collected after the test is similar to this picture below:

The following steps outline the data flow:

• Data Collection: The ESP32 microcontroller continuously collects data from the BNO055 sensor This data includes measurements of roll, pitch, yaw, and acceleration along three axes (Ax, Ay, Az)

Data processing on the ESP32 involves pre-processing the collected data to ensure it is properly formatted, such as timestamped txt files, and ready for storage This stage also includes filtering out errors like noise to maintain data accuracy and reliability.

• Real-Time Data Transmission: Although data is primarily stored locally on the SD card, the ESP32 also supports real-time transmission of data over Wi-

Fi or Bluetooth This allows for remote monitoring and control of the testing environment, ensuring that the system is functioning correctly and that data is being logged as expected

Effective data synchronization is crucial for accurate data acquisition, as it ensures alignment between real-world sensor data and simulation data This is achieved by maintaining consistent timestamps across both systems, enabling precise comparison and analysis of results Proper synchronization enhances the reliability of data-driven insights and improves overall system performance.

Data logging involves saving collected sensor data onto an SD card for offline analysis, ensuring accurate and organized data management The recorded data is structured in columns, including Timestamp (ms), Roll, Pitch, Yaw, Ax, Ay, Az, Gx, Gy, and Gz, facilitating comprehensive analysis of motion and orientation This method enhances data accessibility and supports detailed performance assessments in various applications.

The software component responsible for these tasks is written in C++ using the Arduino framework, providing a straightforward interface for sensor data collection and storage.

AUDEx Integration Workflow

The AUDEx system's integration depends on a seamless data flow between the physical vehicle setup and the simulation environment, ensuring accurate real-time synchronization The process involves several key steps, including data collection from the physical vehicle, transmission to the simulation platform, processing within the simulation environment, and feedback to the vehicle system This smooth workflow enables precise testing and validation of vehicle performance, safety features, and adaptive systems in a controlled, simulated setting Ensuring efficient data exchange between hardware and software components is essential for optimizing the overall functionality of the AUDEx system.

The initial phase of integration focuses on real-time data acquisition from vehicle sensors using the ESP32 microcontroller This data is either logged locally or transmitted to the simulation platform for processing and analysis, enabling precise monitoring and effective vehicle system management.

The simulation setup is carried out using the MANEUVER framework, which is designed to create realistic vehicle behavior models This environment accepts key input parameters such as vehicle mass, tire properties, and road surface conditions to ensure accurate simulation results By leveraging these inputs, the MANEUVER framework generates comprehensive data on vehicle dynamics under specific test scenarios, supporting detailed analysis and performance evaluation.

Real-world vehicle sensor data is compared to simulation results to identify discrepancies between actual and simulated vehicle dynamics This comparison enables precise adjustments to the simulation model, enhancing its accuracy and reliability Accurate data comparison is essential for refining vehicle behavior predictions and ensuring simulation validity.

Model refinement is a crucial step in achieving accurate vehicle simulations After identifying discrepancies, simulation models are iteratively refined to minimize errors This continuous process ensures that the final model accurately reflects real-world vehicle dynamics, resulting in highly reliable and precise simulation results.

Our refined simulation models are rigorously tested across diverse scenarios, including cornering, braking, and acceleration, to ensure their accuracy These evaluations confirm that the models reliably predict vehicle behavior under various driving conditions, enhancing their effectiveness for performance analysis and vehicle design.

Real and Virtual Testing Environments

The AUDEx system combines both real and virtual testing environments to validate vehicle dynamics models

The real testing environment involves conducting physical tests with RC cars fitted with BNO055 sensors and ESP32 data logging systems These tests are performed under controlled conditions on test tracks or closed-course setups to accurately assess vehicle performance during specific maneuvers such as slalom, cornering, and braking.

The Virtual Testing Environment utilizes the MANEUVER framework to run simulations of vehicle dynamics models under consistent conditions This environment enables comprehensive evaluation of a wide range of driving scenarios, including challenging or unsafe situations that are difficult to replicate with physical vehicles By leveraging virtual testing, researchers can improve vehicle safety and performance analysis more efficiently.

The integration of these two testing environments ensures that the vehicle dynamics models are robust and applicable to real-world driving conditions Discrepancies

38 between the two environments are analyzed to refine the simulation models, resulting in a continuous feedback loop that enhances both the real-world and simulated testing phases

Test Objectives and Success Criteria

To demonstrate the methodology effectively, we selected key maneuver tests including the step response, sine sweep, and double lane change These maneuvers encompass a broad range of dynamical behaviors and classification categories, as outlined in the table below By incorporating these diverse maneuvers, the validation process ensures comprehensive coverage of the most critical driving scenarios, confirming the robustness and accuracy of the proposed validation methodology.

The step response maneuver is essential for testing both the steady-state and transient time responses within the linear region of a system As a fundamental open-loop maneuver, it primarily focuses on transient behavior analysis in the time domain This method provides critical insights into system performance and stability, making it a key technique in control system evaluation By analyzing the step response, engineers can effectively assess a system's dynamic response characteristics and improve overall control strategies.

Maneuver Type Input Response Analysis

Step Reponse Fundamental Open Loop Transient Time

Circle Fundamental Open Loop Stochastic Frequency

The subsequent chapters focus on analyzing the selected maneuvers, providing detailed validation metrics and criteria to ensure data accuracy Additionally, the article explains the methods used to effectively handle experimental and simulation data, ensuring robust and reliable results throughout the study.

Table 2: Classification of selected standard test maneuvers according to input type, domain of analysis, application and response type

Response Type Application Domain of

Maneuver Driver (✓) Stochastic Purpose Time Closed Steady-State

- Transient = Response to sudden changes

- Stochastic = Randomized inputs for robustness

- Purpose Dependent = Scenario-based evaluations

- Fundamental = Basic vehicle response tests

- Time Domain = Analysis in time series

- Frequency Domain = Analysis in frequency space

- Closed Loop = With feedback control

- Open Loop = Without feedback control

Verification Testing

Static Analysis

Static analysis is a crucial technique for examining code without execution, enabling the identification of potential errors, enhancing code efficiency, and ensuring adherence to best coding practices This method allows developers to analyze the code structure and quality early in the development process, reducing bugs and improving overall software reliability By thoroughly analyzing the project's code components, we can optimize performance, maintain high code standards, and facilitate easier debugging and future maintenance Implementing static analysis tools is a best practice for promoting high-quality, secure, and compliant software development.

Code Implementation and Explanation below:

• Wire.h: For I2C communication (used by MPU6050)

• SD.h & SPI.h: SD card interface

• MOTION.h: Assumed to handle MPU6050 sensor readings

• SANDISK.h: Assumed to help with SD card operations

• ledPIN: An LED is used as an indicator

• buf_length: Defines the buffer size (holds 5 seconds of data at 100Hz)

• A struct (datapoint) stores timestamp, acceleration (x, y, z), and orientation (roll, pitch, yaw)

• fifoBuf: Circular buffer to store sensor readings

• i: Index to track buffer position

• firstHalfReady & secondHalfReady: Flags for writing data to SD

3 Writing data to SD card

• Opens the SD card file (/car_data_unfiltered.txt) in append mode

• Loops through all datapoint objects in buffer

• Constructs a TXT-style string with: o time

42 o roll, pitch, yaw (formatted with 4 decimal places) o x, y, z (formatted with 3 decimal places)

• Adds a newline (\n) at the end of each row Writes the batch of sensor data to SD card

• Toggles an LED every time data is saved

• Applies calibration offsets to acceleration & gyroscope values

• Computes roll, pitch, yaw (Yaw is estimated using gyroscope integration)

• Handles buffer logic (half-full/full signals data saving)

• Waits for the buffer (firstHalfReady or secondHalfReady)

• Writes half of the buffer at a time to the SD card

• Prevents SD card from being written too often

• Initializes serial, I2C, MPU6050, and SD card

• loop() is empty because FreeRTOS runs everything

The platformio.ini file is essential for configuring your project's environment in PlatformIO, defining the chosen platform, board, and framework to ensure seamless development It also specifies any additional libraries or dependencies needed for your project, streamlining setup and ensuring all necessary components are included Key fields within this configuration file provide clarity and customization, enabling efficient project management and optimized build processes Properly setting up your platformio.ini enhances compatibility, simplifies dependency management, and improves overall development workflow.

The environment configuration name in PlatformIO specifies the setup for a particular development board, such as the ESP32 DevKit V1 This section header within the platformio.ini file indicates which specific environment is being configured, ensuring tailored settings for optimal performance Properly defining this environment is essential for customizing the build process and managing multiple device configurations efficiently.

• env: stands for environment, which defines a build configuration in PlatformIO esp32doit-devkit-v1 refers to the ESP32 DevKit V1 development board (a popular ESP32-based board)

This line sets the baud rate for the serial monitor to 115200

The baud rate is the speed at which data is transmitted over the serial connection between your computer and the ESP32 board A common setting for serial communication is 115200 baud, which offers a good balance of speed and reliability This rate is fast enough for most applications, ensuring efficient data transfer without loss Properly configuring the baud rate is essential for seamless communication with the ESP32 during development and debugging.

This speed is used when you open the serial monitor to view output from your ESP32

The line specifying "espressif32" defines the platform PlatformIO will use for compiling and uploading code to ESP32-based microcontrollers Developed by Espressif Systems, the "espressif32" platform provides all essential tools, drivers, and libraries required for working with ESP32 boards, ensuring seamless development and deployment.

The esp32doit-devkit-v1 identifier specifies the exact ESP32 DevKit V1 board in PlatformIO, guiding the build process to target the correct hardware This configuration is crucial for setting accurate pin configurations, memory layouts, and peripheral support, ensuring optimal performance and compatibility of your project.

The Arduino framework is essential for developing your ESP32 application, offering a user-friendly environment with comprehensive libraries and APIs Arduino, a popular development platform, simplifies hardware interaction, enabling developers to write code efficiently Since the ESP32 utilizes an Arduino core implementation, you can seamlessly use Arduino code and libraries to leverage the device’s advanced capabilities, making development faster and more accessible.

This section lists external libraries that your project depends on PlatformIO will automatically download and include these libraries when you build your project

Here’s the explanation for each library:

The Adafruit BNO055 library enables easy communication with the BNO055 9-axis motion tracking sensor, which combines an accelerometer, gyroscope, and magnetometer to provide accurate orientation data This sensor is ideal for precise motion tracking applications, and the library simplifies integration and utilization of its features in your projects.

The CAN (Controller Area Network) library enables the ESP32 to communicate seamlessly using the CAN bus protocol, which is widely utilized in automotive, industrial, and embedded systems This protocol facilitates efficient data exchange between multiple devices on the same network, making it ideal for reliable and real-time communication in complex environments Implementing the CAN library enhances the ESP32’s versatility, allowing it to seamlessly integrate into various automotive and industrial applications that require robust data transmission.

The SD library enables the ESP32 to seamlessly interface with SD cards, facilitating efficient reading and writing to external storage media This functionality is essential for applications involving sensor data logging, file storage, and data management, making it a vital component for IoT projects requiring reliable data storage solutions.

The SPI (Serial Peripheral Interface) library is essential for enabling communication between the ESP32 and external devices such as sensors, displays, or memory chips using the SPI protocol SPI is a high-speed, synchronous data transfer protocol that facilitates efficient and reliable data exchange By utilizing the SPI library, developers can easily implement communication with various peripherals, making it a vital component in ESP32-based projects.

• Wire: The Wire library is used for I2C communication, another popular communication protocol It allows the ESP32 to communicate with devices that use I2C, such as sensors or displays

The coryjfowler/mcp_can@^1.5.1 library, authored by Cory J Fowler, is essential for enabling communication with MCP2515 CAN Bus modules These modules facilitate seamless integration with CAN networks in embedded systems, allowing for efficient data exchange By specifying version @^1.5.1, you ensure compatibility with version 1.5.1 or later updates, while avoiding potential breaking changes introduced in major releases, thereby maintaining system stability and performance.

The platformio.ini file is essential for configuring your ESP32-based project, defining key environment settings such as board specifications, communication baud rates, and platform details It specifies the use of the ESP32 platform and Arduino framework, ensuring compatibility with various development tools Additionally, the file includes external libraries necessary for sensor integration, SD card operations, and communication modules like CAN bus, enabling seamless hardware interaction and project deployment.

Dynamic Testing Scenarios

In this section information on test vehicle, test track and simulation model are presented

Test vehicle used in the field test is a RC car with 1:8 scale from AUDEx project, HAW Hamburg

Figure 8: RC car in AUDEx project Table 3: RC car technical specifications of the test vehicle

Parameter Description Value Unit g Gravity constant 9.81 m/s² m Vehicle mass 3.884 kg l Wheelbase 0.325 m is Steering ratio 1 - l1 Distance front axle to CG 0.157 m l2 Distance rear axle to CG 0.168 m

Iz Yaw moment of inertia 0.609 kgãm²

Caf Front tyre cornering stiffness 112634 N/rad

Car Rear tyre cornering stiffness 104800 N/rad

The Fzr rear vehicle load measures 18.348 N, with a front slip ratio of 1 minus sigma1 and a rear slip ratio of 1 minus sigma2, indicating the vehicle’s slip behavior during operation Both the normalized front and rear tire cornering stiffness are approximately 5715.4 1/rad and 5714.7 1/rad respectively, reflecting their steering response under load Additionally, the front tire friction coefficient (mu1) is 0.85, contributing to the overall traction and stability of the vehicle's handling characteristics.

E coefficient -1 - mu2 Rear tyre friction coefficient 0.97 -

Table 4 Test Plan for MANEUVER (RC Car)

Name Objective Test Procedure Data to

Analyze vehicle response to a sudden steering input

1 Run straight for 1s (90-degree steering)

2 Apply a sudden fixed steering angle (60 or 120 degrees left/right)

4 Return to straight (90 degree) and hold for 0.3s

6 Motor throttle set up 1500 for 5s

Evaluate vehicle stability and handling in a sudden lane change

1 Accelerate to target speed (throttle

2 Turn left (steering angle 60 degrees) for 0.6s

3 Turn right (steering angle 120 degrees) for 0.6s

4 Return to straight and run for 0.3s

5 Turn right (steering angle 120 degrees) for 0.6s

6 Turn left (steering angle 60 degrees) for 0.6s

7 Return to straight and run for 0.3s

Determine vehicle understeer/ oversteer characteristics

1 Run straight for 5s (90-degree steering, throttle 1550)

2 Gradually apply steering angle (70 degrees left and 110 degrees right), repeat and reverse many times

3 Hold on 3s for each time

4 Increase speed (throttle 1650) and hold for 2s

5 Reduce speed (throttle 1500) and straighten

Validation Testing

Data Analysis

Tiêu đề	Definition, Executing and Evaluating of Verification and Validation Tests (Vehicle Dynamics) for a Vehicle in AUDEx-Project
Tác giả	Trần Đức Bỡnh
Người hướng dẫn	Prof. Dr.-Ing. Dirk Engel, PGS. TS. Trần Quang Vinh
Trường học	Hanoi University of Science and Technology
Chuyên ngành	Vehicle Dynamics
Thể loại	Graduation Project
Năm xuất bản	2025
Thành phố	Hà Nội

Định dạng
Số trang	107
Dung lượng	3,51 MB