2006 ROSES Proposal
Amendment No. 36 to NASA Research Announcement "Research Opportunities in Space and Earth Sciences" (NNH05ZDA001N) establishes a new program element in Appendix A.33 entitled "Advanced Information Systems Technology" (AIST). This new program element solicits proposals for development of selected key technologies to enable an evolution of distributed Earth system sensors and processing components into sensor webs. This AIST program solicitation will concentrate on the architecture (i.e., the design, structure, and behavior) and development of system building blocks leading to autonomous sensor webs. Scenarios are required to show the relevancy of the proposed technology to the objectives of NASA Earth science. During the course of the technology development, the awardees will be required to participate in ESTO-sponsored sensor web technology workshops to advance information sharing on components and concepts. Notices of Intent to propose are due January 17, 2006, and proposals are due March 17, 2006.
- Topic 1: Smart Sensing - Sensor webs of the future may include space-based, airborne, and in-situ sensors, all working together in a semi-closed loop system in which "smart" sensors sense what is happening per their designed sensing capabilities and feed that information into a control system. Based on the sensor inputs, the control system then modifies the environment (instrument pointing, data collection on or off, etc.) and causes the sensors to take in and provide new information to the control system. The system is considered semi-closed because modifications can also be made to the control system by human operators monitoring the sensor web and identified events.
- Topic 2: Sensor Web Communications - The communications component of the sensor web serves to tie the overall system together, forming an infrastructure that allows seamless connectivity among all sensors, nodes, and users belonging to the web. The goal of communication enhancements, especially session layer management, is to support dialog control for autonomous operations involving sensors and data processing and/or modeling entities. Reliable communication links allow the components of the sensor web to move freely within some defined environment while maintaining ubiquitous connectivity and highly reliable data transmission.
- Topic 3:Enabling Model Interactions in Sensor Webs - Technology is sought to tie prediction and forecasting models and scientific analysis tools to the sensor web framework to enable two-way interaction between the modeling / assimilation system and the sensing system to enhance sensor web performance and usage. The goal is to support the creation and management of new sensor web enabled information products.
Since each task is a separate proposal, we need to decide what team (if any) will pursue what task, and then split up into LOI and proposal writing teams. It looks there are two basic areas where we are interested.
Proposal #1: Model-based, smart sensor networks
Team: Burdick, Chandy (PI?), Murray, with help from Doyle, Ho
This proposal would be focused on developing validated methods for targetting sensors in a network to provide information at a given level of confidence.
- If possible, I would like to have Yasi Mostofi (EE postdoc in my group) and perhaps Demetri Spanos (recent CDS PhD) participate in writing this proposal. This will be good experience for the two of them and they can help work up examples and chase down details.
Proposal #2: Robust protocols for sensor web communications
Team: Doyle, Ho (PI?), Low, with help from Chandy, Hassibi?
- Notice of Intent (17 Jan 06) - Notice of Intent (NOI) to propose is encouraged, but not required, for the submission of proposals to this
solicitation. The information contained in an NOI is used to help expedite the proposal review activities and, therefore, is of considerable value to both NASA and the proposer. To be of maximum value, NOIs should be submitted by the PI to NASA’s master proposal data base located at http://nspires.nasaprs.com
- Full Proposal (17 Mar 06) - 12 pages of technical material. $150-500K/year for up to 3 years.
Mani Chandy: most interested in smart sensing, specifically in model-based sensing where the sensing strategy is executed based on sets of models of the environment that is being sensed. This is Topic 1. The ideas are inter-related with Topics 2 and 3 because sensing strategies also depend on communication constraints (Topic 2) and model interactions (Topic 3). Since, however, each proposal has to identify ONE topic, mine is Topic 1.
- Task 1: model based feedback control of active sensors. An issue I'd be interested in is whether there is a general theoretical framework that would allow a more principled approach to the design of protocols.
- Task 2: General design principles for top to bottom protocol stack design for evolvable and robust sensor networks... The idea would be to try to decouple through layering and decentralization as much as possible the hardware and applications levels, so that there could be massive reuse of physical resources for different scientific applications, and natural and transparent upgrade/evolution of capacity/capability.... The issue here is how to allow selected cross-layer interactions as demanded by Task 1, yet preserving the benefits of layered modularity. Sensor networks provide a particular nice challenge in this regard, that the wired Internet generally does not.
Tracy Ho: mostly interested in Task 2, particularly
- autonomous distributed transmission of sensor data* -- distributed adaptation of routing, rate control, data processing, etc. to data rate / correlation structure of sensor data (perhaps Michelle Effros may be interested if compression is considered).
- optimal cost, throughput and reliability trade-offs* -- framework for tuning communication strategy to suit resource constraints or task-related performance requirements; new communication strategies that bring together CS networking concepts, e.g. multipath, zone routing, with info theory concepts, e.g. relay channel, network coding, channel estimation
- distributed storage with latency considerations,
as well as Task 1 on smart sensing, particularly:
- on-demand sensor activity
- failure monitoring, diagnostics and recovery*,
Topics marked with '*' are most closely related to previous or current work.
Richard Murray: I've got ongoing work related to topic #1, as part of some NSF sponsored research in mobile sensor networks. The problem we proposed is to look at how to position sensors to maximize the amount of information they gather. Several students in my group and Joel's group (specifically Tim Chung, Vijay Gupta, Yasi Mostifi) have worked on some initial approaches using mutual information concepts and also (simpler) Kalman filter approaches:
- Communication and Sensing Trade-Offs in Decentralized Mobile Sensor Networks: A Cross-Layer Design Approach
- On a Decentralized Active Sensing Strategy using Mobile Sensor Platforms in a Network
I'm also interested in topic number two and some of Dimitri Spanos' work on robust connectivity might be applicable: