Learning to Rearrange with Physics-Inspired Risk Awareness

TL;DR

This paper presents a reinforcement learning approach for robotic rearrangement tasks that incorporates physics-inspired risk awareness, enabling safer and more efficient manipulation by understanding object properties like mass and friction.

Contribution

The authors introduce a novel physics-inspired reward function and two challenging indoor tasks to improve risk-aware decision-making in robotic manipulation.

Findings

Agents learn policies that minimize physical costs during tasks.

The approach effectively balances task performance and safety considerations.

Policies can be used as safety constraints in critical environments.

Abstract

Real-world applications require a robot operating in the physical world with awareness of potential risks besides accomplishing the task. A large part of risky behaviors arises from interacting with objects in ignorance of affordance. To prevent the agent from making unsafe decisions, we propose to train a robotic agent by reinforcement learning to execute tasks with an awareness of physical properties such as mass and friction in an indoor environment. We achieve this through a novel physics-inspired reward function that encourages the agent to learn a policy discerning different masses and friction coefficients. We introduce two novel and challenging indoor rearrangement tasks -- the variable friction pushing task and the variable mass pushing task -- that allow evaluation of the learned policies in trading off performance and physics-inspired risk. Our results demonstrate that by equipping with the proposed reward, the agent is able to learn policies choosing the pushing targets or goal-reaching trajectories with minimum physical cost, which can be further utilized as a precaution to constrain the agent's behavior in a safety-critic environment.