LVD-NMPC: A Learning-based Vision Dynamics Approach to Nonlinear Model Predictive Control for Autonomous Vehicles

TL;DR

This paper presents LVD-NMPC, a novel learning-based control framework that integrates deep neural networks and vision dynamics to enhance autonomous vehicle navigation and planning.

Contribution

It introduces a new vision dynamics model trained with Deep Q-Learning for improved trajectory prediction in NMPC for autonomous vehicles.

Findings

LVD-NMPC outperforms baseline DWA and PilotNet in simulation and real-world tests.

The approach effectively integrates vision-based dynamics into predictive control.

Enhanced trajectory accuracy and control robustness demonstrated.

Abstract

In this paper, we introduce a learning-based vision dynamics approach to nonlinear model predictive control for autonomous vehicles, coined LVD-NMPC. LVD-NMPC uses an a-priori process model and a learned vision dynamics model used to calculate the dynamics of the driving scene, the controlled system's desired state trajectory and the weighting gains of the quadratic cost function optimized by a constrained predictive controller. The vision system is defined as a deep neural network designed to estimate the dynamics of the images scene. The input is based on historic sequences of sensory observations and vehicle states, integrated by an Augmented Memory component. Deep Q-Learning is used to train the deep network, which once trained can be used to also calculate the desired trajectory of the vehicle. We evaluate LVD-NMPC against a baseline Dynamic Window Approach (DWA) path planning executed using standard NMPC, as well as against the PilotNet neural network. Performance is measured in our simulation environment GridSim, on a real-world 1:8 scaled model car, as well as on a real size autonomous test vehicle and the nuScenes computer vision dataset.