More efficient path optimization for greedy geographic routing in wireless sensor networks

In this paper, we propose an improvement of our previous protocol for geographic routing in wireless sensor networks. The original protocol uses a combination of greedy forwarding and recovery strategies for route discovery. At the same time, shortcut paths are created and then each of these paths is gradually made shorter than the one preceding.

This paper is available online at http://stdb.hnue.edu.vn

MORE EFFICIENT PATH OPTIMIZATION FOR GREEDY GEOGRAPHIC

ROUTING IN WIRELESS SENSOR NETWORKS

Le Dinh Thanh1, Ho Thuan2, Nguyen Dai Tho1

1VNU University of Engineering and Technology

2VAST Institute of Information Technology

1E-mail: thanhld@vnu.edu.vn

Abstract In this paper, we propose an improvement of our previous protocol

for geographic routing in wireless sensor networks The original protocol uses a combination of greedy forwarding and recovery strategies for route discovery At the same time, shortcut paths are created and then each of these paths is gradually

made shorter than the one preceding The scope of destinations is extended from a point to an area The current work further enlarges the destination areas specified

in the routing tables Extensive simulation results show that the upgraded protocol

is more efficient than the previous one

Keywords: Path optimization, geographic routing, wireless sensor networks.

1 Introduction

Geographic routing, a combination of greedy forwarding [1] and recovery routing [2], [7], is particularly suitable to resource-constrained wireless sensor networks Both greedy forwarding and recovery routing can be lightweight in the sense that only information on the position of neighboring nodes is maintained at each node Although existing geographic routing protocols can be very efficient and scalable in dense networks, they generally do not perform well in the presence of holes that represent obstacles to communication

Recently, we proposed GPOR, a greedy with path optimization protocol for geographic routing in wireless sensor networks [11] GPOR deals with the problem of nearly optimal path discovery by using a novel shortcut creation mechanism for early detection and avoidance of obstacles The scope of destinations in the routing tables refers

to an area instead of a point as usually the case in traditional routing The protocol has been shown by simulation to be more efficient than BOUNDHOLE [6] in terms of packet delivery rate and end-to-end delay

In this paper, we improve GPOR by further expanding the destination areas specified in the routing tables Then, extensive simulations are done with the upgraded protocol to compare its performances with GPOR

2 Content

2.1 Related Works

Greedy forwarding [1] is simple, scalable and efficient but suffers from local minima due to obstacles To overcome this problem, greedy forwarding is combined with recovery routing When greedy forwarding fails, a recovery routing strategy is used in order to route the packet to a node where greedy forwarding can be resumed Recovery routing schemes can be classified into four main classes: flooding [2], backtracking [3], face routing [4], [5] and boundary touring [6], [7] Flooding and backtracking are simple but inefficient The main problem with face routing is that existing planar graph extraction methods produce incorrect planar graphs in case of non-uniform radio patterns and thus can cause the routing to fail Boundary detouring is practical and lightweight But it produces long routing paths in cases where packets are forced to detour around large geographic holes In addition, data traffic may become heavy on the boundaries of the holes, something that can lead to congestion

Figure 1 Behavior of nodes in GPOR

The geographic routing protocols in [8] and [9] use a trust-based scheme for early detection and avoidance of obstacles In [10], another path optimization scheme was proposed, based on the marking of areas around obstacles with beacon sequences GPOR [11] (see Figure 1) makes path decisions based on routing tables before switching to greedy forwarding as an option of last resort The routing tables specify shortcut paths

for packet transmission and are updated by a novel shortcut creation technique that makes routing entries better and better after each data transmission Unlike topological routing where each node maintains a routing entry for each destination, each entry in the GPOR routing tables corresponds to a specified destination area This enlargement of the scope

of destinations allows more efficient exploitation of the routing entries To the best of our knowledge, GPOR is the first scheme that can be applied to all-to-all traffic pattern Compared to existing path optimization schemes, GPOR creates shorten paths faster and exploits these paths more efficiently

2.2 Improvement

We further expand the destination areas in the GPOR routing tables by adding a new field, named lastlm, to each table entry The new format of the routing entries is shown in Table 1

Table 1 Extended format of routing entries with lastlm Field Description

pos Targeted position

next Identifier of the neighbor that is chosen as the next hop if the current

routing entry is used

lastlm Position of the last local minimum

ttl Time to live

The expansion of the scope of the destination areas is as follows Given a routing entry hposx, nextx, lastlmx, ttlxi at node C, we define a 2-dimensional Cartesian coordinate system with its origin at lastlmx position and with x-axis being the direction from lastlmxto the targeted position posx We regard the x-axis of the defined coordinate system as the west-east direction (W-E) Then, we divide the plane into four quadrants labeled clockwise NE, ES, SW and WN

The routing entry hposx, nextx, lastlmx, ttlxi at node C is applicable for the destination D if its ttlx is greater than 0 and at least one of the following conditions is met

- Condition 1: The distance from D to posx is not greater than r, where r is the radio range of the nodes

- Condition 2: D is in the ES area, C is either in the WN area or the SW area, and

vector Cnextx points west or D is on the left side of Cnextx if Cnextx points east Figure 2 gives two examples on the scope of applicability of the routing entries

In the first example, vector Cnextx points west, the destination area of the routing entry contains the ES area and those points not farther away from posx than r In the second example, Cnextx points east, the destination area of the routing entry contains the points

in the ES area and on the left of Cnextx, and the points not farther away from posxthan r.

Figure 2 Scope of applicability of the routing entry hposx, nextx, lastlmx, ttlxi

at node C: (a) Cnextx points west; (b) Cnextx points east

2.3 Simulation

We update our GPOR’s implementation in the open-source network simulator ns-2 [12] with the extensions presented in the previous section Then, we perform extensive simulations using the same settings and scripts as in [11] Simulation results show that the extensions to GPOR improve the packet delivery rate and reduce the average path length From now on, we refer to the original version of the GPOR protocol as GPOR-original The simulation results in Figure 3 and Figure 4 show that the packet delivery success rate of GPOR is increased thanks to the expansion of the destination areas The reduction of the average path length as a result of more frequent use of the routing entries

is also confirmed in Figure 5 and Figure 6 Simulation results in Figure 7 and Figure 8 indicate that the expansion of destination areas slightly reduces communication overhead

Figure 3 Packet delivery success rate with varying the number of concurrent traffic flows

Figure 4 Packet delivery success rate with varying network diameter

Figure 5 Average path length with varying the number of concurrent traffic flows

Figure 6 Average path length with varying network diameter

Figure 7 Overhead with varying the number of concurrent traffic flows

Figure 8 Overhead with varying network diameter

3 Conclusion

We have upgraded our previously proposed GPOR protocol with a more efficient path optimization mechanism In the near future, we intend to conduct additional research

on shortcut creation techniques in order to make our geographic routing paradigm even more efficient

REFERENCES

[1] Finn GG., 1987 Routing and Addressing Problems in Large Metropolitan-Scale Internetwork.Tech Rep ISI/RR-87-180

[2] Stojmenovic I, Lin X., 2001 Loop-free hybrid single-path/flooding routing algorithms with guaranteed delivery for wireless networks IEEE Trans on Parallel and Distributed Systems, 12: 1023-1032

[3] Guping Z, Yu W., 2009 Advance detour strategy for geographic routing in wireless sensor networks International Forum on Information Technology and Applications,

pages 296-299.

[4] Karp B, Kung HT., 2000 GPSR: Greedy perimeter stateless routing for wireless sensor networks In Proc of Mobicom, pages 243-254

[5] Bose P, Morin P, Stojmenovic I, Urrutia J., 2001 Routing with guaranteed delivery in

ad hoc wireless networks J Wireless Networks, 7 (6): 609-616

[6] Fang Q, Gao J, Guibas L., 2006 Locating and bypassing routing holes in sensor networks J Mobile Networks and Applications 11 2: 187-200

[7] Nikoletseas S, Powell O., 2007 Simple and efficient geographic routing around obstacles for wireless sensor networks In Proc of the 6th Workshop on Efficient and Experimental Algorithms, pages 161-174

[8] Moraru L, Leone P, Nikoletseas S, Rolim JDP, 2007 Near optimal geographic routing with obstacle avoidance in wireless sensor networks by fast-converging trust-based algorithms In Proc of the 3rd ACM Workshop on QoS and Security for Wireless and Mobile Networks, pages 31-38

[9] Huc F, Jarry A, Leone P, Moraru L, Nikoletseas S, Rolim J., 2009 Early obstacle detection and avoidance for all to all traffic pattern in wireless sensor networks LGOSENSORS, pages 102-115

[10] Koutsopoulos A, Nikoletseas S, Rolim JDP, 2009 Near-optimal data propagation

by efficiently advertising obstacle boundaries In Proc of the 6thACM Symposium on Performance Evaluation of Wireless Ad hoc, Sensor, and Ubiquitous Networks, pages 15-22

[11] Dinh TL, Nguyen DT., 2010 Greedy geographic routing with path optimization in wireless sensor networks In Proc of the 2010 IEEE-RIVF Int’l Conf on Computing and Communications Technologies, pages 148-153

[12] The Network Simulator - ns-2 [Online] Available: http://www.isi.edu/nsna