Fact Probability Vector Based Goal Recognition
Nils Wilken, Lea Cohausz, Christian Bartelt, Heiner Stuckenschmidt
TL;DR
This work tackles online goal recognition by shifting from plan-centric to fact-centric reasoning. It introduces Fact Probability Vector-Based Goal Recognition (FPV), which maps planning states and fact-observation probabilities into a real vector space and computes a distance-based heuristic to rank candidate goals. The method includes an offline probability estimation step via sampling supporter actions from a Relaxed Planning Graph, enabling efficient online recognition. Empirical results across 15 benchmark domains show FPV achieves higher precision, especially under low observability, and substantially better computational efficiency than state-of-the-art baselines, highlighting its practical applicability for real-time goal inference.
Abstract
We present a new approach to goal recognition that involves comparing observed facts with their expected probabilities. These probabilities depend on a specified goal g and initial state s0. Our method maps these probabilities and observed facts into a real vector space to compute heuristic values for potential goals. These values estimate the likelihood of a given goal being the true objective of the observed agent. As obtaining exact expected probabilities for observed facts in an observation sequence is often practically infeasible, we propose and empirically validate a method for approximating these probabilities. Our empirical results show that the proposed approach offers improved goal recognition precision compared to state-of-the-art techniques while reducing computational complexity.