An Oracle-Guided Approach to Constrained Policy Synthesis Under Uncertainty

An Oracle-Guided Approach to Constrained Policy Synthesis Under Uncertainty

Článek recenzovaný mimo WoS a Scopus

Dealing with aleatoric uncertainty is key in many domains involving sequential decision making, e.g., planning in AI, network protocols, and symbolic program synthesis. This paper presents a general-purpose model-based framework to obtain policies operating in uncertain environments in a fully automated manner. The new concept of coloured Markov Decision Processes (MDPs) enables a succinct representation of a wide range of synthesis problems. A coloured MDP describes a collection of possible policy configurations with their structural dependencies. The framework covers the synthesis of (a) programmatic policies from probabilistic program sketches and (b) finite-state controllers representing policies for partially observable MDPs (POMDPs), including decentralised POMDPs as well as constrained POMDPs. We show that all these synthesis problems can be cast as exploring memoryless policies in the corresponding coloured MDP. This exploration uses a symbiosis of two orthogonal techniques: abstraction refinement-using a novel refinement method-and counter-example generalisation. Our approach outperforms dedicated synthesis techniques on some problems and significantly improves an earlier version of this framework.

Dealing with aleatoric uncertainty is key in many domains involving sequential decision making, e.g., planning in AI, network protocols, and symbolic program synthesis. This paper presents a general-purpose model-based framework to obtain policies operating in uncertain environments in a fully automated manner. The new concept of coloured Markov Decision Processes (MDPs) enables a succinct representation of a wide range of synthesis problems. A coloured MDP describes a collection of possible policy configurations with their structural dependencies. The framework covers the synthesis of (a) programmatic policies from probabilistic program sketches and (b) finite-state controllers representing policies for partially observable MDPs (POMDPs), including decentralised POMDPs as well as constrained POMDPs. We show that all these synthesis problems can be cast as exploring memoryless policies in the corresponding coloured MDP. This exploration uses a symbiosis of two orthogonal techniques: abstraction refinement-using a novel refinement method-and counter-example generalisation. Our approach outperforms dedicated synthesis techniques on some problems and significantly improves an earlier version of this framework.

Markov decision processes, model-based reasoning, search, decision making under uncertainty

Markov decision processes, model-based reasoning, search, decision making under uncertainty

MACÁK, F.; ANDRIUSHCHENKO, R.; ČEŠKA, M.; JUNGES, S.; KATOEN, J.

01.02.2025

1076-9757

JOURNAL OF ARTIFICIAL INTELLIGENCE RESEARCH

2025

82

Spojené státy americké

433

469

37

https://www.jair.org/index.php/jair/article/view/16593