MixNet: Structured Deep Neural Motion Prediction for Autonomous Racing

TL;DR

This paper introduces MixNet, a structured deep neural network for predicting opponent vehicle trajectories in autonomous racing, providing accurate forecasts with guaranteed quality, validated through high-fidelity simulations and real-world deployment.

Contribution

The paper presents a novel structured neural network architecture that constrains output trajectories, enabling reliable motion prediction with quality guarantees in autonomous racing.

Findings

Outperforms baseline in prediction accuracy

Maintains quality guarantees in predictions

Successfully deployed on racing vehicle in real-world scenario

Abstract

Reliably predicting the motion of contestant vehicles surrounding an autonomous racecar is crucial for effective and performant planning. Although highly expressive, deep neural networks are black-box models, making their usage challenging in safety-critical applications, such as autonomous driving. In this paper, we introduce a structured way of forecasting the movement of opposing racecars with deep neural networks. The resulting set of possible output trajectories is constrained. Hence quality guarantees about the prediction can be given. We report the performance of the model by evaluating it together with an LSTM-based encoder-decoder architecture on data acquired from high-fidelity Hardware-in-the-Loop simulations. The proposed approach outperforms the baseline regarding the prediction accuracy but still fulfills the quality guarantees. Thus, a robust real-world application of the model is proven. The presented model was deployed on the racecar of the Technical University of Munich for the Indy Autonomous Challenge 2021. The code used in this research is available as open-source software at www.github.com/TUMFTM/MixNet.