Information Theory for Mobile Ad-Hoc Networks (ITMANET): The FLoWS Project Thrust 2 Overview: Layerless Dynamic Networks Lizhong Zheng

Layerless Dynamic Networks Dynamic and Interference: out main challenges –Lack of channel/network side information, high costs of coordination overhead, non-ergodic/atypical behavior; –Interference suppression, cancellation, and utilization; –High level relation between a dynamic problem and a interference problem? Separation of functionalities with different time scale no longer valid Layerless: our vehicle to break away from conventional approaches –Network information theory: heterogeneous data and soft information processing, how much information survives through the network? –Structured codes: with good error exponent, flexibility for joint processing, combining and relaying, new interface to the physical layer; –Interference design and management: broadcasting channel, cooperative and cognitive networking; –Operating with imperfect side information, robustness and universal designs.

Intellectual Tools and Focus Areas Dynamic Network Information Theory: beyond point-to-point communications –Relay, soft (lossy) information processing; –Generalized network coding: putting fragments of information together; –New approach to utilize feedbacks, manage error and ARQ; –Broadcasting, Interference and cooperation; Structured coding: concrete designs for cross layer processing –Joint source/channel/network coding; –Flexible partial decoding for weak channels; –New metric to interface and control the physical layer; Assumptions and Robustness: networking without perfectly reliable/precise side information –Universal algorithms requires less coordination; –Tradeoff between cooperation and coordination;

Thrust Achievements Dynamic Network Information Theory Relay, forwarding and combining soft information –Likelihood forwarding– Koetter –General relaying for multicast – Goldsmith –DMT for multi-hop networks -- Goldsmith Optimal end-to-end rate-reliability tradeoff from dynamic decode-and forward relays; Interference forwarding, utilizing the structure of codebook for relay processing; Feedback, channel memory and dynamics –Degraded Finite State Broadcasting Channel – Goldsmith –Feedback and directed information in Networks – Goldsmith –Error-Erasure tradeoff for compound channels – Moulin Erasure/ARQ, and bursty traffic hard to measure by average performance; Universal sense of confidence for compound channel, generalizing classical results.

Thrust Achievements Dynamic Network Information Theory Generalized network coding –General capacity using network coding – Medard –Matroidal solutions – Effros & Keotter Interference, cooperation and coordination –Euclidean Information Theory – Zheng –Cognitive interference channel – Goldsmith –“Exponential family” type of Broadcasting channel – Coleman New family of “E-type” channel defined, with capacity region of broadcasting channel (not necessarily degraded) computable; Direct application in network security and new mode of information embedding;

Thrust Achievements Structured coding Joint source/channel/network coding with layered codes –Broadcasting with layered source codes – Goldsmith –Generalized capacity/distortion for joint source channel codes – Effros & Goldsmith –Joint S/C coding in networks – Coleman –Diversity-Distortion Tradeoff – Medard & Zheng Dynamic view of distortion over channel, allowing graceful degradation of the reconstruction quality in outage; Instead of long term average performance, explicitly address the tradeoff in opportunistic operation; Layered code designs –Message embedding with UEP – Zheng Performance limits of protecting critical partial information; Using critical message to initiate retransmissions; Natural connections to feedback channel;

Thrust Achievements Robustness and Side Information Robust algorithms designs –Universal decoding with N-P setup – Moulin –Mismatched decoder – Meyn, Zheng, Medard –Error erasure exponents tradeoff for compound channels– Moulin –Additive universal receiver for compound channel – Zheng Using additive metric allows existing encoder/decoder to be reused for compound channels; Geometric insights to generalized linear receivers offers new view of universal receiver; Plug-n-play polytope receiver with well controlled performance- complexity tradeoff Utilizing limited control/feedback links –Feedback in wireless networks – Goldsmith –Joint source channel coding for CSI feedbacks

Goldsmith: general relaying, soft combining Goldsmith: Cognitive users and interference Zheng: Embedded Coding and UEP Moulin: Information flow via timing Goldsmith: Broadcasting with layered code Coleman: Joint Source/Channel Coding in Networks Dynamic Network Information Theory CSI, feedback, and robustness Structured coding Effros, Goldsmith: Generalized capacity, distortion, and joint source/channel coding. Thrust 2 Achievements Overview Zheng: Euclidean Information Theory Goldsmith: Degraded FS Broadcast Channels Moulin: Universal Decoding in MANETs Goldsmith: Feedback and Directed Information Coleman: Rate Distortion of Poisson Processes Moulin: Error/erasure tradeoff for compound channel Goldsmith: DMT for multi-hop networks Coleman: “E-type” broadcasting channels Zheng: Message embedding in feedback channels Medard, Zheng: Diversity-distortion tradeoff Goldsmith: Interference forwarding

Provide building blocks for large networks, translate design constraints into network modeling assumptions Thrust Context Thrust 2 Dynamic Network Information theory: improving performance in presence of interference, cooperation, and dynamic environment Thrust 3 Application Metrics and Network Performance Guide problem formulation by identifying application constraints and relevant performance metrics Provide achievable performance region, based on which distributed algorithms and resource allocation over large networks are designed Thrust 1 New Paradigm of outer bounds Performance benchmark and design justification

Roadmap for meeting Phase 2 Goals Evolve results in all thrust areas to examine more complex models, robustness/security, more challenging dynamics, and larger networks. –Efficient coordination and cooperation over the network; –New way of communication with layered coding, over broadcasting or interference channels, soft(lossy) combining relaying strategy; –Universal and robust algorithm, misinformed nodes; Demonstrate synergies between thrust areas –Cross layer designs, joint source-channel-network coding, with new end- to-end performance metrics; –Heterogeneous data processing, distinguishing different types of data, different types of error, and different types of services; –Dynamic networks with time variation and adaptation; Demonstrate that key synergies between information theory, network theory, and optimization/control lead to at least an order of magnitude performance gain for key metrics. –New interface to the physical layer, measuring the tradeoff between different types of data and the cost of network protocols; –Building blocks for network utility maximization, scalable and distributed networking;