Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper tackles the challenge of scaling solutions to the Optimal Observability Problem (OOP) within Partially Observable Markov Decision Processes (POMDPs), focusing on the Sensor Selection Problem (SSP) and Positional Observability Problem (POP). They introduce a novel POMDP decomposition method to identify effective observation functions. The new approach achieves a 3-order-of-magnitude improvement in instance size and a 5-order-of-magnitude improvement in runtime compared to prior parameter synthesis techniques.
Decomposing POMDPs unlocks orders-of-magnitude speedups in solving sensor selection problems, making observation-aware planning practical for larger, more complex uncertain domains.
Deciding which sensing capabilities to deploy on an agent in uncertain domains is a fundamental engineering challenge, in which one balances task achievability against the high costs of hardware and processing. This problem has previously been formalized as the Optimal Observability Problem (OOP), based on the well-known Partially Observable Markov Decision Process (POMDP) model for decision-making. This work studies (sub-)symbolic techniques to scale solving of decidable fragments of the OOP, namely the Sensor Selection Problem (SSP) and the Positional Observability Problem (POP). Besides improving the original approach based on parameter synthesis, we develop a new solving method that identifies sensible observation functions via decomposition of POMDPs, improving performance by 3 and 5 orders of magnitude for instance size and runtime, respectively.