Multiuser Receiver Aware Multicast in CDMA-based Multihop Wireless Ad-hoc Networks Parmesh Ramanathan Department of ECE University of Wisconsin-Madison Madison, WI Kuang-Ching Wang Department of ECE University of Wisconsin-Madison Madison, WI
Multihop Wireless Ad-hoc Network Computing devices with wireless communication capabilities have become commercially available – e.g., laptops, PDA’s, or specialized sensor devices. Without infrastructure support, these devices (nodes) form an ad-hoc network. Based on a cooperative assumption, nodes forward messages for neighbors when requested, and a message potentially has to traverse multiple hops to reach its destination. Nodes can be mobile, and rely on limited battery life.
Multicast Applications in MANET Multicast performs one-to-many data dissemination. Multicast best facilitates the persistent need of data dissemination to a selected set of nodes in the network. Potential applications are: Wireless classroom/conference: a lecturer/speaker multicasts texts, images, or multimedia contents to selected audience. Wireless battlefield: a commander multicasts commands, graphical information, or real- time images to selected personnel on the field. Wireless sensor network: a node multicasts data archives, computation results to selected nodes to perform collaborative signal processing.
Challenges Power Optimal coverage and tree topology Multiuser interference Mobility Nodes join and leave – tree grows/shrinks Nodes move – tree changes Tree robustness Throughput Scheduling Multiple channels for transmission Scalability Cost of maintenance (tree computation) Interference level
Existing Solutions Decoupled network layer protocols No assumption made of the underlying link and physical layers Pros: Clean and portable implementations Cons: Inefficient usage of bandwidth and power Can be augmented with topology control schemes Maximum transmit power adjustment based on topology Channel variations and multiuser interference not accounted for Joint network and link/physical layer approaches Energy efficient trees for DS-CDMA networks, Ephremides et al. Build trees with optimized total transmit power for static wireless network Estimate required transmit power based on distance law Multiuser interference not considered
What is missing? Multiuser interference: Major issue in wireless communication which can not be overlooked. Interference characteristics are determined by the multiuser receiver. Adaptive tree adjustment: Physical channels are changing, as well as node locations. Centrally generate a new tree periodically Locally adjust sub-tree structure periodically Adaptive power control To cope with random fading channels To cope with node mobility
Problem Statement Given a set of nodes with specified multiuser receiver path gains, and random noise plus background interference model Find a multicast tree rooted at the source, and the transmit power of all non-leaf nodes such that: (i) the SINR requirements of all nodes are satisfied (ii) the total transmit power is minimized.
Random Fading and Mobility Node-to-node path gain changes constantly due to random fading effects of the wireless channel, and node mobility. Power control at two granularities: Random fading effects are in short time scales A local fast power control scheme should be adopted. Mobility causes large path-gain changes. Tree adjustment can be needed to maintain power optimality.
Tree and Power Adjustment Tree adjustment can be done in both centralized and distributed manners. Centralized: Pros: better optimized in power Cons: computationally expensive Distributed: Pros: only local exchange of information needed Cons: tree structure can be far from optimal Power adjustment must be fast in response to channel variations. Centralized allocation is unrealistic. Distributed power control algorithm is well-known.
Proposed Schemes Finding a tree optimized in total transmit power is NP-hard. Two heuristics are proposed: Centralized-Tree Distributed Power multicast scheme (CTDP) Distributed-Tree Distributed Power multicast scheme (DTDP) Transmit power estimates are based on realistic modeling of the multiuser receivers. CTDP periodically recomputes a new tree using a centralized incremental power heuristic, adding one node at a time with the least incremental transmit power. In DTDP, each node locally decides on its position in the tree by collaborating with nearby nodes. Distributed power control is used in both schemes to account for path- gain variation due to both random fading and mobility.
Simulation Results (1/2) Simulations of CTDP and DTDP are carried out in MATLAB ® with three different multiuser receivers: matched filter, minimum mean squared error, and decorrelator receivers. Using CTDP, different optimal trees are obtained when different multiuser receivers are assumed. Significant power savings are achieved in optimizing with the assumed multiuser receiver model.
Simulation Results (2/2) In spite of the computational efficiency, DTDP is less successful than CTDP in finding the optimal tree. This is expected due to the lack of global knowledge and less flexibility in tree changing scope. The SINR distribution for CTDP and DTDP both demonstrate their satisfactory achievement in meeting all nodes’ SINR requirements. (Left) the average per node transmit power vs. time and (right) SINR distribution for CTDP and DTDP with MMSE receivers.