A Data-Efficient Approach to Precise and Controlled Pushing

TL;DR

This paper demonstrates that a data-efficient, model-based control approach using Gaussian processes can achieve precise pushing manipulation with minimal training data, emphasizing control over explicit system modeling.

Contribution

Introduces a data-efficient control method for pushing tasks by learning dynamics with Gaussian processes and applying model predictive control, reducing data requirements significantly.

Findings

Achieves complex pushing trajectories with only 10 data points.

Uses Gaussian process models within a model predictive control framework.

Effectively handles actuator and task constraints in planar manipulation.

Abstract

Decades of research in control theory have shown that simple controllers, when provided with timely feedback, can control complex systems. Pushing is an example of a complex mechanical system that is difficult to model accurately due to unknown system parameters such as coefficients of friction and pressure distributions. In this paper, we explore the data-complexity required for controlling, rather than modeling, such a system. Results show that a model-based control approach, where the dynamical model is learned from data, is capable of performing complex pushing trajectories with a minimal amount of training data (10 data points). The dynamics of pushing interactions are modeled using a Gaussian process (GP) and are leveraged within a model predictive control approach that linearizes the GP and imposes actuator and task constraints for a planar manipulation task.