Technical program - (in [pdf] format)

Session I: Power Control & Scheduling (8:30-10:00 am)

  • On asymptotic optimality of dual scheduling algorithm in a generalized switch
    Lijun Chen, Steven Low, John Doyle (California Institute of Technology)
  • Optimal power allocation and scheduling for two-cell capacity maximization
    Anders Gjendemsjo (Norwegian University of Science and Technology), David Gesbert (Institute Eurecom), Geir E. Oien (Norwegian University of Science and Technology), Saad G. Kiani (Institute Eurecom)
  • Scheduling algorithms for point to multipoint operation in IEEE 802.16 networks
    Rajagopal Iyengar, Koushik Kar, Biplab Sikdar (Rensselaer Polytechnic Institute)

Coffee break (10:00-10:30 am)

Keynote talk (10:30-11:30 am): Randall Berry (Northwestern University)
Wireless Resource Allocation Games (Abstract)

Invited talk (11:30 am-12:10 pm): Saswati Sarkar (University of Pennsylvania)
Distributed and Partial Information based Optimal Control in Multi-Hop Wireless Networks (Abstract)

Lunch break (12:10-1:20 pm)

Session II: Ad-hoc Networks (1:20-2:20 pm)

  • Distributed Allocation of Identical Resources in Mobile Ad Hoc Networks
    Salahuddin Mohammad Masum (Daffodil International University), Amin Ahsan Ali (University of Dhaka)
  • Alert service in VANET: Analysis and design
    Roberta Fracchia, Michela Meo (Politecnico di Torino)

Invited talk (2:20-3:00 pm): Aditya Karnik (University of Waterloo, Canada)
Throughput-optimal Configuration of Wireless Sensor Networks (Abstract)

Joint work with Catherine Rosenberg and Ravi R. Mazumdar, (University of Waterloo, Canada)

Coffee break (3:00-3:30 pm)

Session III: IEEE 802.11 Networks (3:30-5:30 pm)

  • A self-managed distributed channel selection algorithm for WLANs
    D.J. Leith, P. Clifford (Hamilton Institute, National University of Ireland)
  • On-line client-AP association in WLANs
    Gaurav S. Kasbekar, Pavan Nuggehalli, Joy Kuri (Indian Institute of Science Bangalore)
  • Improving fairness in multi-hop mesh networks using 802.11e
    K. Duffy, D.J. Leith, T. Li, D. Malone (Hamilton Institute, National University of Ireland)
  • Queue management strategies to improve TCP fairness in IEEE 802.11 wireless LANs
    Mingwei Gong, Qian Wu, Carey Williamson (University of Calgary, Canada)

Abstracts of the keynote and invited talks

Wireless Resource Allocation Games (Randall Berry, keynote talk)

Game theory provides a useful set of tools for studying distributed resource allocation. In this talk, we discuss some game theoretic models for resource allocation in wireless networks. As an example, we focus on the case where a set of nodes is sharing a common spectrum band and the primary resource to be allocated is the node's transmission power. We give a distributed resource allocation protocol whch only relies on limited information exchange. In certain cases this is shown to achieve optimal performance. (Joint work with Michael Honig and Jianwei Huang at Northwestern.)

Distributed and Partial Information Based Optimal Control in Multihop Wireless Networks (Saswati Sarkar, invited talk)

One of the main challenges in optimally scheduling transmissions in multihop wireless networks is that each node has limited information both about its neighbors and about its own resources such as transmission channels. In this talk, first, we will assume that each node has limited information about its neighbors, but has full information about the states of all its channels. We will consider a simple distributed scheduling strategy, maximal scheduling that can be used in this scenario, and prove that it attains a guaranteed fraction of the maximum throughput region in arbitrary wireless networks. Furthermore, the guarantees are tight in that they can not be improved any further with maximal scheduling. Next, we will assume that a node incurs a cost whenever it acquires information about the states of its channels. Owing to these costs, the amount of information the node acquires about its channels becomes an important decision variable. Our goal now is to determine a jointly optimal transmission and information acquisition strategy that maximizes a utility function that depends on both the performance and the cost accrued in probing the channels for learning their states. We will present policies that approximate the optimal solution within guaranteeable constant factors. Finally, we will present several open problems in this research area.

Throughput-optimal Configuration of Wireless Sensor Networks (Aditya Karnik, invited talk)

In this work we seek answers to two fundamental questions concerning the data gathering wireless sensor networks; first, for a given placement of $n$ sensors and the sink what is the maximum achievable throughput of the network?, and second, how should the network, i.e., the radio and link layer parameters at each sensor be configured to achieve this maximum? Unlike the popular ``scaling'' approach, we determine what is achievable but not through asymptotic results. We assume centrally computed TDMA link schedules and not a distributed MAC. We show that routing and scheduling are intricately related. Hence, we cast the problem of maximizing the network throughput as a nonlinear nonconvex optimization problem over the radio parameters (transmission power and modulation), routing and scheduling schemes. In a special case of fixed transmission power and modulation scheme, we show that the optimal throughput is determined by the maximum weighted clique of the contention graph prescribed by the radio parameters; the vertex weights in this graph equal the traffic carried by the corresponding link under the routing scheme that is optimal for power $P$. Moreover, the optimal link schedule is contention free and is also determined by the maximum weighted clique. For a grid topology with the sink in a corner, and all the sensors using the same radio parameters, we obtain the maximum throughput in a closed form under a two-hop interference model. The optimal routing is such that in a certain region around the sink the traffic is routed using the shortest paths while the traffic outside this region flows in two branches deviating away from each other and finally getting fed into the region through the border sensors. Interestingly, of all feasible transmission powers, the power which allows sensors to transmit to the sink in one hop has the maximum throughput. Joint work with Catherine Rosenberg and Ravi R. Mazumdar, (University of Waterloo, Canada)