Research
My research interests are in multivehicle control applied to Lagrangian distributed sensing applications. Specifically, my Ph.D. research at UC Berkeley has focused on building a fleet of floating, actuated water quality sensors for studying riverine and estuarine systems; see float.berkeley.edu for further information about the floating sensor fleet.
The motivation for the floating sensor devices is to investigate Lagrangian sensing: in normal operation, the vehicle doesn't use its propulsion, and so is carried along by the natural river current. As it moves though the estuary or delta, its path is a source of useful information.
There are several reasons why augmenting the Lagrangian sensor with actuation is a good idea:
- Safety, logistics: The real world is an inconvenient place. Without actuation, passive sensors in constrained environments like rivers tend to run into the banks, obstacles, etc. Passing boats often throw a wake that pushes small surface vehicles to the bank. With actuation, the vehicles can rescue themselves from some of these hazards; this improves the overall fleet performance.
- Distribution: In a delta with many junctions, there may be regions where the coverage probability is low; in other words, the channel would eventually get visited by a passive sensor, but it might take a long time. Actuated vehicles can be controlled to maximize the coverage over the delta, by changing their trajectory through junctions to take the "less travelled" branch.
- Active sensing: Active sensing is a non-Lagrangian mode. Nevertheless, there are many possibilities for an actuated sensor that changes its trajectory in response to its collected sensor data. One application would be to track the movement of a concentration isoline through the estuary.