Real-time simulation and control of high-definition matrix headlights for automated driving

Abstract

Modern matrix headlights with up to 1.3 million controllable light sources called pixellights enable precise illumination control and sophisticated lighting functions for automated driving. However, their development, verification and validation present substantial challenges that require real-time headlight simulation and control capabilities to be mastered efficiently in terms of time and cost. This thesis at hand addresses these challenges through three primary contributions: a novel real-time matrix headlight simulation for commercial rendering engines, an intuitive and flexible real-time matrix headlight control algorithm and a comprehensive matrix headlight rapid prototyping and development tool. The developed methods have been extensively validated by testing with real matrix headlights. The first significant contribution is a physically accurate, real-time capable simulation approach that combines standard rendering engine spotlights with Sparse Matrix-Vector Multiplication to be lighting technology-independent. This method achieves real-time performance while maintaining physical accuracy by leveraging established sparse matrix formats and parallel computing algorithms. The second significant contribution is the SuperSampling Control (SSC) algorithm for real-time matrix headlight control of arbitrary lighting functions and beam patterns. SSC synthesizes ray tracing, Supersampling Anti-Aliasing and the MapReduce parallel processing method into an intuitive yet computationally efficient approach for rapidly prototyping arbitrary dynamic lighting functions. The third significant contribution is developing a matrix headlight development tool, which integrates the proposed real-time simulation and control methods. The software supports hybrid development approaches through virtual rapid prototyping and hardware-in-the-loop testing with real matrix headlights. Practical applications are energy-efficient, environment-optimal illumination and observer-specific distortion-free symbol projection with closed-loop feedback control.