SensorWeb Data Fusion in Large Arrays of Microsensors MURI Review Meeting Introduction/Overview Alan S. Willsky September 22, 2003

The Day’s Agenda Introduction/Overview Research Presentations Summary

Introduction/Overview Outline Setting the stage: Intellectual themes and research concentration areas Facts, Statistics, and Such What we have heard from the EAB/TAC Outline of remaining presentations and how they fit

Intellectual Themes IT-1: Consistent (or manageably inconsistent) fusion of networked, myopic sensors IT-2: Fusion of heterogeneous sensors in unstructured and uncertain environments IT-3: Wireless networks, network communication and information theory

Research Concentration Areas RCA-1: (Self-)Calibration RCA-2&3: Fundamental limits on fusion, network information theory, tradeoffs in local vs global processing RCA-4: Bounds & characteristics of algorithms to ID minimum resources needed to detect, estimate, track… RCA-5: Fusion Algorithms RCA-6: Distributed Algorithms with guarantees on global behavior (both positive & negative!) RCA-7: Create events/data for experiments and demos

Funding, Timing, and Staffing We are 3 years, 1 month into this program Funding: We are on plan Includes adjustments for delays in funding and other ARO- approved variations (e.g,. Sanjoy Mitter’s time this fall) Principal Faculty/Investigators Willsky, Mitter, Jaakkola, Fisher, Cetin, Tsitsiklis (MIT); Kulkarni, Verdu (Princeton); Kumar (Illinois) Numerous post-docs, visitors, and grad students MIT (19), Illinois (12), Princeton (12)

Website: Continually growing record of accomplishments and activities The Website:

Publications, Presentations, Results Publications Latest count: 113 publications Includes a (growing) number of key invited journal papers Presentations Numerous internal and external talks Continually growing record of invited, plenary, and keynote talks Continuing participation in Battlefield Acoustics mtg. Significant research results in all aspects of our intellectual agenda See talks to follow, publications, summary at the end, website

A few of our industrial/DoD interactions With PARC (formerly Xerox PARC) Joint Paper, Transition and Experimental Demo within SensIT program, Maurice Chu has joined PARC team With ALPHATECH Transition of fusion methods to several programs (Raptor, DARPA DTT program (with SenTech), DARPA ISP Program) ARO SBIR Topic A MURI research (Tsitsiklis) cited as foundational to the topic With ARL: Numerous since last review Visits to ARL by Verdu, Kumar Battlefield Acoustics Mtg. (Willsky-Sadler, Pham, Srour, Moses) Interactions with Sensors CTA (Falco, Moses) Visits to MIT: Sadler, Moses Extended follow-ups with ARL: Ihler, Fisher, Cetin, Malioutov

Sampling of Academic Interactions Berkeley/Stanford MURI Post-doctoral positions of Drs. Sekhar Tatikonda and Martin Wainwright Continuing interchange and reciprocal visits MURI on Battlefield Visualization Interaction with Prof. Pramod Varshney Ohio State University Collaboration with Prof. Randy Moses Through Sensors CTA Reciprocal visits (Moses, Fisher) Utilization of our network-constrained estimation algorithms in Prof. Moses’ work Several collaborative research projects initiated

Standing Questions from the EAB/TAC - I Time spent on the project We are maintaining/exceeding levels anticipated (Mitter 4 mo., Willsky 3 mo., Kulkarni 2/3 mo., Kumar 2/3 mo., Jaakkola 2/3 mo., Verdu 2/3 mo.) Cost-sharing Substantial for faculty (Schools pay almost all AY time) Even greater for students (More than twice as many students involved as are receiving direct support) Plans for handling real data, for demos and/or paths to implementation Variety of examples given in previous reviews Further discussions embedded in talks this year

What we heard from the EAB/TAC last year - I Significant praise Caliber of research Conscientiousness of our efforts to meet MURI objectives and respond to suggestions Several issues/requests made Allow more time for discussion and questions Fewer talks with more time between for informal discussions during breaks Expand interactions with ARL See preceding list plus discussions embedded in today’s presentations Enhance collaboration among investigators Talks today will make that collaboration more apparent

What we heard from the EAB/TAC last year - II Issues/requests continued Make sure to consider communication constraints in the context of fusion objectives as well as limitations of low power Several talks address problems related to this Is “optimal” processing required if it implies global communications; can’t one achieve adequate performance with more “local” fusion? See later talks Make sure work is relevant to sensor modalities of interest to the Army One of the purposes of our interactions with ARL Also, see several talks today

Agenda - I Welcome/Overview 8:30 – 9:15 Lavery, Willsky: Welcome/Overview 9:15 – 9:50 Cetin: A Sparse Signal Reconstruction Perspective for Source Localization with Sensor Arrays IT-2, RCA-1 (also ties to RCA-4) 9:55 – 10:30 Fisher (Kulkarni/Verdu): Information-Theoretic Approaches to Data Association and Fusion in Sensor Networks IT-2, RCA-1,5 10:30 – 10:45 Break 10:45 – 11:20 Willsky (Kumar): Network-Constrained Estimation IT-1, RCA-4,5,6 (also ties to RCA-2&3)

Agenda - II 11:20 – 1:00 Lunch 1:00 – 1:35 Jaakkola (Tsitsiklis): Using Heterogeneous Data Sources and Sequential Resource Management IT-1, RCA-4,5,6 1:40 – 2:15 Kumar (Verdu/Tsitsiklis): Information Transfer in Wireless Networks for Distributed Sensing and Control IT-2, RCA-2&3 2:15 – 2:30 Break

Agenda - III 2:30 – 3:10 Mitter/Chu/Willsky: Distributed Multiple Target Tracking and Information Architecture for Designing Applications on Ad Hoc Sensor Networks IT-1, RCA-4,5,6 3:10 – 3:30 Willsky: Summary 3:30 – 4:30 Government Caucus 4:30 – 5:00 Preliminary Debrief