Optimizing Non-Markovian Information Gain Under Physics-Based Communication Constraints
A recent paper by members of DCIST proposes an exploration method that maintains communication between all robot team members and a static base station. By maintaining communication while exploring, robots are kept up to date on the progress of other team members and important information—e.g., survivors in a search and rescue mission—are quickly transmitted to a static base station. Their method uses a combination of optimization and sampling to quickly find paths for each robot that maintains communication, then optimizes the paths to maximize the information gain relative to the total path cost. Their method is verified using a realistic communication model and obtains 2-5 times more information relative to a path’s cost than other state of the art works.
Capability: T3C2E: Adversarial network and motion synthesis
Points of Contact: Neil Dantam (PI), Qi Han, John Rogers, and Matthew Schack
Citation: M. A. Schack, J. G. Rogers, Q. Han, and N. T. Dantam, “Optimizing non-markovian information gain under physics-based communication constraints,” IEEE Robotics and Automation Letters, vol. 6, no. 3, pp. 4813–4819, 2021.