Dynamic Error Sources and Real-Time Compensation Methods in Multi-Sensor Fusion Systems for Autonomous Vehicles

This paper analyzes the sources of dynamic errors and the coupling propagation mechanism from the perspectives of sensor systems and the environment, thus clarifying their important impacts on the fusion performance. It further clearly reviews two main methodological approaches, namely the explicit modeling based on physical principles and the implicit modeling driven by data, and also constructs a real-time compensation technology framework from perception-front-end preprocessing to the collaborative optimization of the fusion process. This paper also carries out a comparative evaluation of the performance trade-offs among different methods. Dynamic errors have multiple sources, time-varying characteristics and nonlinearity, which then requires balancing and selecting compensation strategies by weighing model interpretability, computational efficiency and robustness in specific scenarios. Physical models have clear theoretical frameworks, while data-driven approaches show considerable potential in capturing complex error patterns.