Randomized 3D Position-based Routing Algorithms for Ad-hoc Networks A.E. Abdallah, T. Fevens and J. Opatrny Department of Computer Science and Software Engineering Concordia University Montr´ eal, QC, Canada, H3G 1M8 Email: {ae abdal,fevens,opatrny}@cse.concordia.ca Abstract In position-based routing algorithms for ad-hoc net- works, the nodes use the geographical information to make the routing decisions. Recent research in this field primar- ily addresses such routing algorithms in two dimensional space (2D). However, in real applications, nodes may be distributed in 3D space. In this paper we extend previous randomized routing algorithms from 2D space to 3D space, and we propose two new position-based routing algorithms that combine randomized AB3D routing algorithms with a deterministic CFace (coordinate face) algorithm. The first algorithm AB3D-CFace(1)-AB3D starts with AB3D routing algorithm until a local minimum is reached. The algorithm then switches to CFace routing using one projected coordi- nate. If CFace(1) enters a loop, the algorithm switches back to AB3D. The second algorithm AB3D-CFace(3) starts with AB3D, until a local minimum is reached. The algorithm then permanently switches to CFace routing using three projected coordinates, in order. We evaluate our mech- anisms and compare them with the current routing algo- rithms. The simulation results show the significant improve- ment in delivery rate over pure AB3D randomized routing (97% compared to 70%) and reduction in path dilation (up to 50%) over pure CFace algorithm. 1. Introduction Mobile ad-hoc networks (MANETs) consist of a collec- tion of wireless mobile hosts that can communicate with each other without a fixed infrastructure. A node in the network can communicate directly only with its neighbors (the nodes within its transmission range). To communi- cate with nodes outside its transmission range, multihop routing is used utilizing intermediate communicating nodes. Since mobile ad-hoc networks may change their topology frequently and because of the resource constraints, routing in such networks is difficult. In the past decade, several adaptive routing protocols for ad-hoc networks have been proposed to address the multihop routing problem in ad- hoc networks. Each is based on different assumptions and concepts. In general, these protocols can be classified in two basic types: topology based routing and position-based routing. Topology based routing protocols define an explicit route among nodes using the information about the links that exist in the network. Position-based routing [1, 2, 3, 4, 5, 6, 10] or online rout- ing [9, 16] algorithms limit the huge bandwidth required by topology based routing. The host forwards the message based on its position, the position of the destination, and the position of the hosts to which it can communicate directly. In one class of position-based routing, progress-based algo- rithms, the current node forwards the packet in every step to exactly one of its neighbors, which is chosen accord- ing to some heuristic such as Greedy [5] or Compass [4]. However, progress-based routing methods suffer from the so-called local minimum phenomenon, in which a packet may get stuck at a node that does not have a neighbor that makes a progress to the destination, even though the source and destination are connected in the network. Many algo- rithms attempt to deal with this problem. Bose et al. [9] described a routing algorithm which guarantees the deliv- ery of the message in a MANET under a geometric planar graph. This algorithm, called Face routing, uses the right hand rule to propagate the packet along the interior of the faces of the planar graph which are intersected by the line segment connecting the source to the destination. A com- bination between Greedy routing and Face routing has been proposed in [9], GFG (Greedy-Face-Greedy), and in [11], GPSR (Greedy Perimeter Stateless Routing). Initially, these algorithms make greedy forwarding decisions. If the packet reaches a region where progress to the destination by greedy forwarding is impossible, the algorithm enter into recovery mode by switching to face routing. Once the packet reaches a node closer to the destination than that node where greedy forwarding previously failed for that packet, the algorithm switches back to greedy forwarding again. Face routing al- gorithms require a planar subgraph to guarantee the deliv- 1-4244-0499-1/06/$20.00 ©2006 IEEE