Energy-constrained Active Exploration with Incremental Perception

Active Exploration Framework

An active exploration framework that optimizes target search under energy constraints given no prior information. We consider a robot with an incremental-resolution symbolic perception deployed in an environment with known geometry and unknown semantics. Casting this energy-constrained sequential decision-making problem as a flow maximization problem, we leverage tools from optimization and graph theory to efficiently plan online as the robot’s knowledge about the semantics of the surroundings is updated.

Disha Kamale
Disha Kamale
PhD Student, Robotics

My research interests include safe, perception-aware decision-making and planning with user preferences.