IEEE Communications Magazine • March 2005 101 0163-6804/05/$20.00 © 2005 IEEE TOPICS IN AD HOC NETWORKS WHAT IS THE TRANSMISSION RADIUS? The designers of network layer protocols for ad hoc and sensor networks assume the unit disk graph (UDG) communication model, where two nodes communicate if and only if they are within distance R, where R is the transmission radius, equal for all nodes. Almost all articles even use R as the independent variable in their simulations. While the protocols at the network layer are designed with simple assumptions and perfor- mance metrics, experiments are normally carried on simulators that implement more realistic phys- ical and medium access control (MAC) layers. Simulators are trying to match the physical layer, which suggests that the UDG model is not realistic because it ignores random variations in received signal strengths. It was demonstrated that signal strength fluctuations have a signifi- cant impact on ad hoc network performance metrics, sometimes “outperforming” the impact of node mobility. Thus, nondeterministic radio fluctuations cannot be ignored when designing robust ad hoc network protocols based on ad hoc network simulation and analysis. Assuming fixed signal-to-noise ratio (SNR), the model used in simulators and that hopefully matters in real equipment then looks like the one in Fig. 1, which shows how packet reception probability p(x) depends on distance x between two nodes. The exact shape of the curve depends on the exact model used (combined Friis and two-ray ground model in [1]; lognormal shadow- ing model [2, 3]). It is obvious that the UDG model is indeed a good initial approximation for this, since the reception probability is close to 0 or 1 everywhere except around the transmission radius. But what is the transmission radius in Fig. 1? Is R = 30 (refer to distances in Fig. 1), meaning that the failure rate for transmissions is ≤ 5 percent? Is R = 50, meaning that the recep- tion probability is ≥ 5 percent (i.e., two nodes Ivan Stojmenovic, Amiya Nayak, and Johnson Kuruvila, SITE, University of Ottawa ABSTRACT We present guidelines on how to design net- work layer protocols when the unit disk graph (UDG) model is replaced by a more realistic physical layer model. Instead of merely using the transmission radius in the UDG model, physical, MAC, and network layers share the information about a bit and/or packet reception probability as a function of distance between nodes. We assume that all nodes use the same transmission power for sending messages, and that a packet is received when all its bits are correctly received. The MAC layer reacts to this probabilistic reception information by adjusting the number of acknowledgments and/or retrans- missions. We observe that an optimal route dis- covery protocol cannot be based on a single retransmission by each node, because such a search may fail to reach the destination or find the optimal path. Next, we discuss that gaining neighbor knowledge information with “hello” packets is not a trivial protocol. We describe localized position-based routing protocols that aim to minimize the expected hop count (in case of hop-by-hop acknowledgments and fixed bit rate) or maximize the probability of delivery (when acknowledgments are not sent). We pro- pose a guideline for the design of greedy posi- tion-based routing protocols with known destination locations. The node currently hold- ing the message will forward it to a neighbor (closer to the destination than itself) that mini- mizes the ratio of cost over progress, where the cost measure depends on the assumptions and metrics used, while the progress measures the difference in distances to the destination. We consider two basic medium access layer approaches, with fixed and variable packet lengths. This article will serve as a preliminary contribution toward the development of net- work layer protocols that will match the assump- tions and criteria already used in simulators and ultimately in real equipment. Design Guidelines for Routing Protocols in Ad Hoc and Sensor Networks with a Realistic Physical Layer