Fast Model Identification via Physics Engines for Data-Efficient Policy Search

TL;DR

This paper introduces a data-efficient method for robot model identification using physics engines and Bayesian optimization, enabling effective policy learning with fewer real-world experiments.

Contribution

It proposes a strategy to identify a sufficiently accurate model for policy optimization, reducing the need for precise model reproduction and minimizing real-world trials.

Findings

Significantly improves data efficiency in policy search.

Effective in both simulation and real robotic tasks.

Reduces number of real-world experiments needed.

Abstract

This paper presents a method for identifying mechanical parameters of robots or objects, such as their mass and friction coefficients. Key features are the use of off-the-shelf physics engines and the adaptation of a Bayesian optimization technique towards minimizing the number of real-world experiments needed for model-based reinforcement learning. The proposed framework reproduces in a physics engine experiments performed on a real robot and optimizes the model's mechanical parameters so as to match real-world trajectories. The optimized model is then used for learning a policy in simulation, before real-world deployment. It is well understood, however, that it is hard to exactly reproduce real trajectories in simulation. Moreover, a near-optimal policy can be frequently found with an imperfect model. Therefore, this work proposes a strategy for identifying a model that is just good enough to approximate the value of a locally optimal policy with a certain confidence, instead of wasting effort on identifying the most accurate model. Evaluations, performed both in simulation and on a real robotic manipulation task, indicate that the proposed strategy results in an overall time-efficient, integrated model identification and learning solution, which significantly improves the data-efficiency of existing policy search algorithms.