A Universal Vehicle-Trailer Navigation System with Neural Kinematics and Online Residual Learning

2025-07-21

AI Summary

Autonomous navigation of vehicle-trailer systems—particularly those incorporating castor-wheel trailers—faces significant challenges in modeling and generalization across diverse payloads and operating conditions. Method: This paper proposes a hybrid modeling framework integrating neural kinematics with online residual learning. It synergistically combines classical vehicle kinematics with a lightweight neural network to model trailer dynamics, enabling automatic adaptation to varying trailer configurations and dynamic disturbances without manual calibration. Furthermore, a weighted model-combination scheme is incorporated into a model predictive control (MPC) framework to enhance long-horizon prediction accuracy and path safety. Contribution/Results: Extensive experiments across multiple real-world trailer configurations and load conditions demonstrate that the system achieves high-precision navigation and stable control without manual parameter tuning. It exhibits strong robustness, superior cross-scenario generalization, and effective real-time error correction.

Technology Category

• Intelligent Robots: State Estimation
• Planning, Routing, and Scheduling: Model-Based Reasoning
• Multiagent Systems: Multiagent Learning

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization

Abstract

Autonomous navigation of vehicle-trailer systems is crucial in environments like airports, supermarkets, and concert venues, where various types of trailers are needed to navigate with different payloads and conditions. However, accurately modeling such systems remains challenging, especially for trailers with castor wheels. In this work, we propose a novel universal vehicle-trailer navigation system that integrates a hybrid nominal kinematic model--combining classical nonholonomic constraints for vehicles and neural network-based trailer kinematics--with a lightweight online residual learning module to correct real-time modeling discrepancies and disturbances. Additionally, we develop a model predictive control framework with a weighted model combination strategy that improves long-horizon prediction accuracy and ensures safer motion planning. Our approach is validated through extensive real-world experiments involving multiple trailer types and varying payload conditions, demonstrating robust performance without manual tuning or trailer-specific calibration.

Problem

• Accurate modeling of vehicle-trailer systems with castor wheels
• Real-time correction of kinematic model discrepancies and disturbances
• Improving long-horizon prediction and motion planning safety

Innovation

• Hybrid kinematic model with neural networks
• Online residual learning for real-time correction
• Weighted model predictive control framework