Increasing Connection Lifetimes through Dynamic Distribution of Budgeted Power Zeki Bilgin Computer Science Department Graduate Center, City University of New York (CUNY) Email: zbilgin@gc.cuny.edu Bilal Khan Department of Mathematics & Computer Science John Jay College, CUNY Email: bkhan@jjay.cuny.edu Abstract—We present a new dynamic scheme which continu- ously redistributes a fixed power budget among mobile wireless nodes participating in a multi-hop wireless connection, with the objective of maximizing the expected lifetime of the connection. Our experimental simulations indicate that the proposed power budget distribution scheme yields a significant increase in con- nection lifetime. We quantify the sensitivity of the performance gains to various system parameters, including connection size, node density, power budget size, and mean node velocities. We then compare the efficacy of our scheme with two schemes: one that simply distributes the power budget uniformly, and one that distributes the power budget dynamically with the objective of minimizing end-to-end bit error rate (BER). In comparing the relative performance of the three schemes we obtain a qualitative assessment of the inherent oppositions and tradeoffs between the objectives of BER minimization and lifetime maximization. I. I NTRODUCTION Historically [1] reconciling the gap between power con- sumption and supply involved solving the following issues: (i) improving the power efficiency in the system; and (ii) pre- venting the system deconstruction due to unfair power usage. In their earlier work [2], [3] the authors proposed addressing these concerns by normalizing the measurement of relative “efficiency” and “fairness” using a model in which every connection is assigned a fixed power utilization budget. This assigned budget reflects the connection’s priority, or equiva- lently, the benefit that the system derives in maintaining the connection. In consumer MANETS, for example, this benefit might be based on financial incentives provided by a paying satisfied customer, while in military MANETs it could reflect the extent to which the connection is essential to achieving a positive outcome for some coordinated mission objective. In [4], the authors considered the opportunities afforded by such a model vis-a-vis minimizing connection bit error rate (BER), and presented a distributed scheme which minimized a connection’s end to end BER by continuously reapportioning its power budget among its constituent (static) nodes. This work diverges and extends the earlier investigations of the authors [4] in two very significant ways: First, this paper considers mobile nodes instead of merely considering static snapshots of a dynamic network; secondly, our objective here is to leverage the ability to dynamically distribute a connection’s power budget towards maximizing expected con- nection lifetime, rather than towards minimizing connection BER. We will compare our proposed lifetime-maximizing scheme with the connection lifetimes enjoyed by the BER- minimizing power distribution scheme of [4], and in doing so obtain a sense of the extent to which the two objectives are in opposition. II. RELATED WORK Efficient power management for MANETs has been in- vestigated in prior research at several protocol layers (see e.g. [1]). As an objective, lifetime maximization has been interpreted in one of two ways: network lifetime maximization, and connection lifetime maximization. Network Lifetime. The lifetime of a network is most frequently defined as the time interval for which the network is a connected graph. Broadly speaking, the network may partition (becoming disconnected) when one of two events occurs: (i) the autonomous movement of a node causes some of its incident link(s) to fail due to a shortage of transmission power, or (ii) some node exhausts its energy supply sufficiently so that some of its incident links fail. Most prior research on network lifetime attempts to delay the onset of these two types of events—the most frequent emphasis being on event type (ii), see e.g. [5], [6]—by extending the network routing protocol to make it energy-aware and using a route selection strategy that facilitates optimization with respect to the network’s lifetime. Connection Lifetime. Somewhat analogously, a connec- tion’s lifetime is typically taken to be the time interval during which all of the connection’s constituent links are operational. A link in a connection ceases to be operational when one of two events occurs: (a) the autonomous movement of one of the link’s endpoints causes it to fall out of transmission range of the other endpoint, or (b) one of the two endpoints exhausts its energy supply, causing the other endpoint fall out of transmission range. Most prior research on connection lifetime attempts to delay the onset of these two types of events— the most frequent emphasis being on events of type (a), see e.g. [7], [8], [9], [10], [11]. The main approach has been (as was the case in research on network lifetime) to extend the network routing protocol by making it energy-aware, and then to make route selection sensitive to connection lifetime maximization. In [7], [8], [9], [10], for example, the authors proposed new routing protocol extensions based on finding the path which probably has longest lifetime among many possible 978-1-4244-2202-9/08/$25.00 © 2008 IEEE