ARTFEED — Contemporary Art Intelligence

KAPS: Exact Algorithm for Minimax Regret in Uncertain MDPs with Small Policy Sets

ai-technology · 2026-08-04

Researchers introduced k-adaptable policy synthesis, a new approach for optimizing a set of k policies under a minimax-regret objective in uncertain Markov decision processes (UMDPs). The problem is proven NP-hard, and the team developed KAPS, an exact nested branch-and-bound algorithm. This work addresses real-world sequential decision-making where model uncertainty is resolved shortly before execution, allowing selection of the most suitable policy from a limited prepared set. The research is available on arXiv with ID 2608.02509.

Key facts

  • The paper is titled 'Optimizing Minimax Regret in Uncertain MDPs with Small Sets of Policies'.
  • It is available on arXiv with ID 2608.02509.
  • The research introduces k-adaptable policy synthesis.
  • The objective is to minimize maximum regret (minimax regret).
  • The problem is proven to be NP-hard.
  • The authors developed an exact algorithm called KAPS.
  • KAPS uses a nested branch-and-bound approach.
  • The setting involves model uncertainty resolved shortly before execution.

Entities

Institutions

  • arXiv

Sources