Throughput Analysis of an Aloha-Based MAC Policy for Ad Hoc Networks Konstantinos Oikonomou, Member, IEEE, and Ioannis Stavrakakis, Senior Member, IEEE Abstract— Re-use of existing widely explored Medium Access Control (MAC) schemes, like the well-known Aloha scheme, is not applicable in ad hoc networks where the transmissions of the users can be normally sensed by only a fraction of the users present in the network. Therefore, estimations of the network traffic load are not possible anymore. Here, an adaptive probabilistic policy for medium access control in ad hoc networks, inspired by the Aloha paradigm, is proposed and analyzed. Simulation results show that this policy is capable of achieving higher system throughput when compared to other policies that have been proposed for ad hoc networks. It is also shown that mobility severely impacts the system throughput and therefore, an alternative approach is proposed that reduces the effects of mobility in the expense of the maximum achievable system throughput. Index Terms— Ad Hoc, Aloha, MAC. I. I NTRODUCTION The design of Medium Access Control (MAC) policies in ad hoc networks is challenging due to the idiosyncratic behavior of these networks. Several MAC policies have been proposed, [1], [2], [3], [4], [5], which are based on the CSMA/CA mechanism, including in most of the cases the Ready-To-Send/Clear-To-Send handshake dialogue to avoid the hidden/exposed terminal problem. TDMA-based MAC protocols have also been proposed (e.g., [6]) and it has been shown that when an optimal solution is required, the derivation of the scheduling (time slots in which a node is allowed to transmit during a frame), is an NP-complete problem, similar to the n-coloring problem in graph theory, [7], [8]. Consequently, these approaches are not suitable for ad-hoc networks where, in general, nodes are moving and therefore, the scheduling needs to be recalculated for all nodes in the network. TDMA-based MAC policies, which do not require recal- culation of the scheduling of the nodes when the topology of the network is changing and the frame size is significantly smaller than the number of nodes in the network, have already been proposed, [9], [10], [11], [12]. The Deterministic Policy (referred to hereafter as D-Policy) has originally been pro- posed in [9]. Under this policy nodes are allowed to transmit only at a (small) subset of the available time slots carefully selected so that at least one of them be collision free. While the latter results in a guaranteed minimum throughput per node, This work has been supported in aprt by the E-NEXT research program that is partly funded by the European Commission. restricting the transmission opportunities of a node to a (small) subset of the available slots, leads to a fairly low overall system throughput, [12]. Since most of the non-assigned - under the D-Policy - slots may be wasted if other nodes are temporarily idle or move away, it has been proposed in [12] that such slots be utilized probabilistically. This is the key idea behind the Probabilis- tic Policy (referred to hereafter as P-Policy) introduced in [12]. It turns out that the system throughput is (in general) significantly increased under the P-Policy. The higher system throughput under the P-Policy is achieved by giving access to all nodes to all slots, with probability 1 if the slot is assigned (under the D-Policy) to a node and with access probability p otherwise. Certainly, the idea of probabilistic transmission attempts is not new. It is influenced by related work done in the seventies (e.g., Aloha, [13]) for shared medium environments with common channels. However, in ad hoc networks even for the cases that the medium is shared (e.g., wireless environments) no common channel for all users is in general present. As it is depicted in Figure 1, nodes are not, in general, aware about the transmissions of all other nodes in the network and estimations of the backlogged traffic are not possible, as opposed to the case of a bus network architecture for which the family of the Aloha protocols is destined for. 0 2 5 3 4 1 0 5 3 4 1 2 0 5 3 4 1 2 (a) (b) (c) Fig. 1. Examples of shared medium networks. (a) Bus Network Architecture and (b) Fully Connected Network: it is possible for all nodes to sense the transmission of any other node in the network. In general, in Ad Hoc Networks, some nodes may not be able to sense the transmission of other nodes as it is the case depicted in (c). For example, in (c) it is not possible for node 5 to sense any transmission of node 0, while in the special case of (b) it is. Even though the work carried out for the family of the Aloha protocols cannot be reused due to the aforementioned