Answer-Set-Programming-based Abstractions for Reinforcement Learning

In three linesNew approach combining Answer-Set Programming (ASP) and Reinforcement Learning to create logical abstractions of state spaces. Authors reimplement the CARCASS framework (originally in Prolog) using ASP, a fully declarative language, and evaluate it on Blocks World and Minigrid. ASP provides richer modelling for logical representations of Markov Decision Processes.

Summary generated by Claude — human-verified