I've Changed My Mind: Robots Adapting to Changing Human Goals during Collaboration

TL;DR

This paper presents a novel robot adaptation method for dynamic human goals during collaboration, enabling robots to detect goal changes and actively assist in complex, real-world tasks like cooking.

Contribution

It introduces a goal change detection technique using multiple candidate sequences and a policy bank, improving robot responsiveness to evolving human objectives.

Findings

Outperforms baseline goal prediction algorithms in collaborative cooking tasks.

Reduces task completion time after human goal switches.

Enhances collaboration efficiency in dynamic goal scenarios.

Abstract

For effective human-robot collaboration, a robot must align its actions with human goals, even as they change mid-task. Prior approaches often assume fixed goals, reducing goal prediction to a one-time inference. However, in real-world scenarios, humans frequently shift goals, making it challenging for robots to adapt without explicit communication. We propose a method for detecting goal changes by tracking multiple candidate action sequences and verifying their plausibility against a policy bank. Upon detecting a change, the robot refines its belief in relevant past actions and constructs Receding Horizon Planning (RHP) trees to actively select actions that assist the human while encouraging Differentiating Actions to reveal their updated goal. We evaluate our approach in a collaborative cooking environment with up to 30 unique recipes and compare it to three comparable human goal prediction algorithms. Our method outperforms all baselines, quickly converging to the correct goal after a switch, reducing task completion time, and improving collaboration efficiency.