Real-time Neural-MPC: Deep Learning Model Predictive Control for Quadrotors and Agile Robotic Platforms

TL;DR

This paper introduces Real-time Neural MPC, a framework that enables the use of large neural network models for dynamics in real-time control of agile robots, significantly improving accuracy and complexity handling.

Contribution

The work presents a novel method to efficiently incorporate large neural network dynamics models into real-time MPC for embedded systems, enabling higher model complexity and better control.

Findings

Achieved real-time control with neural networks over 4000 times larger in capacity.

Reduced positional tracking error by up to 82% compared to traditional MPC.

Demonstrated successful deployment on a real quadrotor platform.

Abstract

Model Predictive Control (MPC) has become a popular framework in embedded control for high-performance autonomous systems. However, to achieve good control performance using MPC, an accurate dynamics model is key. To maintain real-time operation, the dynamics models used on embedded systems have been limited to simple first-principle models, which substantially limits their representative power. In contrast to such simple models, machine learning approaches, specifically neural networks, have been shown to accurately model even complex dynamic effects, but their large computational complexity hindered combination with fast real-time iteration loops. With this work, we present Real-time Neural MPC, a framework to efficiently integrate large, complex neural network architectures as dynamics models within a model-predictive control pipeline. Our experiments, performed in simulation and the real world onboard a highly agile quadrotor platform, demonstrate the capabilities of the described system to run learned models with, previously infeasible, large modeling capacity using gradient-based online optimization MPC. Compared to prior implementations of neural networks in online optimization MPC we can leverage models of over 4000 times larger parametric capacity in a 50Hz real-time window on an embedded platform. Further, we show the feasibility of our framework on real-world problems by reducing the positional tracking error by up to 82% when compared to state-of-the-art MPC approaches without neural network dynamics.