Abstract: In the adaptive information gathering problem, a robot is required to select an informative sensing location using the history of measurements acquired thus far. While there is an extensive amount of prior work investigating effective practical approximations using variants of Shannon’s entropy, the efficacy of such policies heavily depends on the geometric distribution of objects in the world. On the other hand, the principled approach of employing online POMDP solvers is rendered impractical by the need to explicitly sample online from a posterior distribution of world maps. We present a novel data-driven imitation learning framework to efficiently train information gathering policies. The policy imitates a clairvoyant oracle – an oracle that at train time has full knowledge about the world map and can compute maximally informative sensing locations. We analyze the learnt policy by showing that offline imitation of a clairvoyant oracle is implicitly equivalent to online oracle execution in conjunction with posterior sampling. This observation allows us to obtain powerful near-optimality guarantees for information gathering problems possessing an adaptive submodularity property. As we demonstrate on a spectrum of 2D and 3D exploration problems with aerial vehicles, the trained policies enjoy the best of both worlds – they adapt to different world map distributions while being computationally inexpensive to evaluate.

In closely related very recent work, I will highlight how to leverage submodularity to plan efficient trajectories for 3D reconstruction which are environment and battery constraint-aware resulting in better reconstructions for the constant battery budget, compared to the state-of-the-art.

Bio:  Debadeepta Dey is a researcher in the Adaptive Systems and Interaction (ASI) group led by Dr. Eric Horvitz at Microsoft Research, Redmond. He received his PhD at the Robotics Institute, Carnegie Mellon University, advised by Prof. Drew Bagnell. He is interested in bridging the gap between perception and planning for autonomous robots. His interests include decision-making under uncertainty, reinforcement learning, planning and perception. Nowadays he is especially interested in aerial vehicle autonomy ranging from small quadrotors to large gliders and balances his time between fundamental theoretical advances as well as pushing the state-of-the-art in realizable practical systems.

Homepage:  www.debadeepta.com