|
| Research | Projects | Testbed | Publications |
|
RESEARCH
|
Data Management
The MIST group is working on the design of various query processing and data
management interfaces for sensor networks. The goals of this work
are to design high-level "macroprogramming" interfaces that allow users
to efficiently query a sensor network without worrying about low-level
details related to power management, communications scheduling, and
ad-hoc network maintenance.
We are also looking at issues focusing on the use of modeling techniques to allow
users to ask queries over
noisy, lossy, and uncertain data that is endemic to sensors and other
real-world data sources. In particular, we are looking at
probabilistic models as a way to capture historical correlations
between sensors which can then be used to infer missing values
and approximately answer queries using just a subset of readings.
Sensor Network Protocols
Our research in this area is driven by challenges unique to sensor networks, such as short-range wireless networking,
communication and power budgets, and need for maintenance-free operation. We investigate
the issues in congestion control, ad hoc routing, robustness, management and low-power operation of sensor networks.
Applications
Our research enables the development of novel sensor network applications, but also
addresses new problems exposed in the course of their development. Our work on indoor location and node localization supports
a new class of location-aware applications. We are also developing a car-based sensor network as a platform
for mobile sensor network research and a variety of real-world applications useful to drivers.
|
PROJECTS
|
Fusion
Mitigating Congestion in Wireless Sensor Nets
- Congestion degrades the wireless channel
- Result: wasted energy and unfairness
- Solution: hop-by-hop flow control, rate limiting, MAC techniques
- Combination: Fusion
- Bret Hull, Kyle Jamieson and Hari Balakrishnan. "Mitigating Congestion in Wireless Sensor Networks" ACM SenSys 2004. [PDF]
|
Wavescope
System Support for High-Rate Applications
- High data-rate sensornet applications: industrial monitoring, structural monitoring (pipelines, bridges), medical sensors
- Processing 100Hz - 10KHz samples: embedded signal processing, adaptive model-based information processing, efficient wireless traffic management
|
Cartel
Cars as Sensor Data Sources
- Car sensor monitoring and diagnostics
- Collaborative traffic monitoring/mapping
- Intermittent connectivity: local storage and opportunistic 802.11 access to share data
- Peer-to-peer information sharing
|
Memento Mori
Distributed Failure Detection
- Efficient delivery of coarse-grained health alerts
- Robust failure detection
- Distributed, load-balanced logging of symptoms
- Remote symptom analysis; time, cause of failure
|
Cricket
Indoor Location System
- Indoor location system for pervasive and sensor-based computing environments
- Fine-grained location information for handhelds, laptops, and sensor nodes
- Commercially available from Crossbow
- Visit the Cricket website for more information
|
MIT CSAIL Testbed
Experimental Sensornet Infrastructure
- Mica2/Cricket hybrid deployment (Telos coming soon).
- Ethernet backchannel for experimental data collection.
- Perl-based job scheduling tools.
- MistLab website
|
PUBLICATIONS
|
- Bret Hull, Kyle Jamieson and Hari Balakrishnan. "Mitigating Congestion in Wireless Sensor Networks" ACM SenSys 2004. [PDF]
|
