The Autonomous Systems and Controls Laboratory focuses on fundamental research in control theory and autonomous systems.
Limited communication and too much noise
How much communication does a task force of vehicles really need to autonomously complete a given mission? The answer is surprisingly small. Dan Stilwell and his students in the Autonomous Systems and Controls Laboratory (ASCL) are developing a new class of algorithms that not only enable teams of autonomous vehicles to cooperate, but do so with as little communication as possible between vehicles.
The researchers are not just developing algorithms, but also testing and perfecting them using Virginia Tech’s own fleet of Autonomous Underwater Vehicles (AUVs). The group recently field tested data fusion and motion control algorithms using two Virginia Tech 475 AUVs. Each AUV towed a custom hydrophone array built by Virginia Tech students that measures the relative bearing to a source of acoustic noise. The AUVs maneuvered themselves into the best positions to cooperatively identify the location of an acoustic source. During the entire mission, the AUVs exchanged only a total of seven one-way data packets.
The lab has also developed a new class of extended/iterated Kalman filters that correctly account for state-dependent sensor noise and show significant performance improvement over more typical Kalman filters. These filters are useful for many applications, but were developed to help with the hydrophone array towed by the lab’s AUVs. The noise of the sensor depends on the state of the system and this violates the requirements of a typical Kalman filter. “We are very excited that we have actually implemented these ideas on real vehicle systems,” Stilwell said. “It is rare and very exciting for new theoretical developments such as ours to be demonstrated in the field, and to have the field trials so convincingly demonstrate the utility of the new theory!”