Adaptive Informative Path Planning with Multimodal Sensing

Abstract

Adaptive Informative Path Planning (AIPP) problems model an agent tasked with obtaining information subject to resource constraints in unknown, partially observable environments. Existing work on AIPP has focused on representing observations about the world as a result of agent movement. We formulate the more general setting where the agent may choose between different sensors at the cost of some energy, in addition to traversing the environment to gather information. We call this problem AIPPMS (MS for Multimodal Sensing). AIPPMS requires reasoning jointly about the effects of sensing and movement in terms of both energy expended and information gained. We frame AIPPMS as a Partially Observable Markov Decision Process (POMDP) and solve it with online planning. Our approach is based on the Partially Observable Monte Carlo Planning framework with modifications to ensure constraint feasibility and a heuristic rollout policy tailored for AIPPMS. We evaluate our method on two domains: a simulated search-and-rescue scenario and a challenging extension to the classic RockSample problem. We find that our approach outperforms a classic AIPP algorithm that is modified for AIPPMS, as well as online planning using a random rollout policy.

Sequential Bayesian Optimization for Adaptive Informative Path Planning with Multimodal Sensing

“…Section IV formulates the AIPPMS problem as a belief MDP and describes our approach for solving the problem. Section V presents our results on two AIPPMS problems and compares our method to Choudry et al and other baseline methods [1]. Section VI concludes our work.…”

“…2) The extensive evaluation of our method in simulation, along with directly comparing to previous AIPPMS methods. We compare our results on the Information Search RockSample problem introduced by He et al [17] and adapted by Choudry et al [1]. We also introduce a new benchmark for AIPPMS problems: the Rover Exploration problem.…”

“…In addition, the agent has limited time and energy resources requiring that it balance exploring the environment and exploiting its knowledge of the environment based on its current belief. This problem is referred to as Adaptive Informative Path Planning with Multimodal Sensing (AIPPMS) [1]. The objective of the AIPPMS problem is to produce an adaptive decision making strategy to guide the agent to the goal location while balancing exploration with exploitation and respecting the resource constraints.…”

“…The adaptive problem extends the IPP problem by seeking a policy that changes in response to received observations [3][4][5][6][7]. The AIPPMS problem further extends the adaptive problem by reasoning over multiple sensing modalities with different sensing costs corresponding to sensor accuracy and movement costs [1].…”

“…This extension adds additional complexity by requiring that the agent reasons about the costbenefit trade-off associated with the different sensing modalities. Choudry et al considered multiple sensing modalities [1]. We build upon this work by formulating the problem as a belief Markov decision process (MDP) as shown in Fig.…”