Online Pareto-Optimal Decision-Making for Complex Tasks using Active Inference

TL;DR

This paper presents a novel multi-objective reinforcement learning framework for robots that balances safety, task objectives, and user preferences using active inference, demonstrating superior performance in complex, uncertain environments.

Contribution

It introduces a two-layer framework combining multi-objective planning with active inference for decision-making, improving safety, adaptability, and user alignment in robotic tasks.

Findings

Outperforms existing methods in learning multiple optimal trade-offs

Successfully adheres to user preferences in decision-making

Enables user adjustment of trade-off balances

Abstract

When a robot autonomously performs a complex task, it frequently must balance competing objectives while maintaining safety. This becomes more difficult in uncertain environments with stochastic outcomes. Enhancing transparency in the robot's behavior and aligning with user preferences are also crucial. This paper introduces a novel framework for multi-objective reinforcement learning that ensures safe task execution, optimizes trade-offs between objectives, and adheres to user preferences. The framework has two main layers: a multi-objective task planner and a high-level selector. The planning layer generates a set of optimal trade-off plans that guarantee satisfaction of a temporal logic task. The selector uses active inference to decide which generated plan best complies with user preferences and aids learning. Operating iteratively, the framework updates a parameterized learning model based on collected data. Case studies and benchmarks on both manipulation and mobile robots show that our framework outperforms other methods and (i) learns multiple optimal trade-offs, (ii) adheres to a user preference, and (iii) allows the user to adjust the balance between (i) and (ii).