Accelerated gradient descent for high frequency Model Predictive Control

TL;DR

This paper investigates the effectiveness of first-order gradient descent methods in high-frequency Model Predictive Control for robotics, demonstrating they can match the performance of more complex second-order methods.

Contribution

The study shows that first-order methods can be a viable and simpler alternative to second-order methods in high-frequency MPC for robotics.

Findings

First-order methods can achieve comparable performance to second-order methods.

First-order methods simplify implementation in real-time robotic control.

Potential for more efficient MPC algorithms in robotics applications.

Abstract

The recent promises of Model Predictive Control in robotics have motivated the development of tailored second-order methods to solve optimal control problems efficiently. While those methods benefit from strong convergence properties, tailored efficient implementations are challenging to derive. In this work, we study the potential effectiveness of first-order methods and show on a torque controlled manipulator that they can equal the performances of second-order methods.