Online Planning in POMDPs with State-Requests

In key real-world problems, full state information is sometimes available but onlyat a high cost, like activating precise yet energy-intensive sensors or consulting hu-mans, thereby compelling the agent to operate under partial observability. For thisscenario, we propose AEMS-SR (Anytime Error Minimization Search with StateRequests), a principled online planning algorithm tailored for POMDPs with staterequests. By representing the search space as a graph instead of a tree, AEMS-SRavoids the exponential growth of the search space originating from state requests.Theoretical analysis demonstrates AEMS-SR’s ε-optimality, ensuring solution qual-ity, while empirical evaluations illustrate its effectiveness compared with AEMSand POMCP, two SOTA online planning algorithms. AEMS-SR enables efficientplanning in domains characterized by partial observability and costly state requestsoffering practical benefits across various applications.

Journal article