ODGR: Online Dynamic Goal Recognition

TL;DR

This paper introduces ODGR, a new approach for real-time recognition of an agent's changing goals using reinforcement learning, addressing scalability and dynamic goal challenges in goal recognition tasks.

Contribution

It formulates the novel problem of Online Dynamic Goal Recognition, incorporating dynamic goals into GR and demonstrating its feasibility with transfer learning in navigation domains.

Findings

Feasibility of solving ODGR in navigation domains.

Reformulation of existing approaches using ODGR.

Potential for robust real-time goal recognition in dynamic environments.

Abstract

Traditionally, Reinforcement Learning (RL) problems are aimed at optimization of the behavior of an agent. This paper proposes a novel take on RL, which is used to learn the policy of another agent, to allow real-time recognition of that agent's goals. Goal Recognition (GR) has traditionally been framed as a planning problem where one must recognize an agent's objectives based on its observed actions. Recent approaches have shown how reinforcement learning can be used as part of the GR pipeline, but are limited to recognizing predefined goals and lack scalability in domains with a large goal space. This paper formulates a novel problem, "Online Dynamic Goal Recognition" (ODGR), as a first step to address these limitations. Contributions include introducing the concept of dynamic goals into the standard GR problem definition, revisiting common approaches by reformulating them using ODGR, and demonstrating the feasibility of solving ODGR in a navigation domain using transfer learning. These novel formulations open the door for future extensions of existing transfer learning-based GR methods, which will be robust to changing and expansive real-time environments.