Infrastructure sensor-based cooperative perception for early stage connected and automated vehicle deployment

Abstract

AbstractInfrastructure-based sensors provide a potentially promising solution to support the wide adoption of connected and automated vehicles (CAVs) technologies at an early stage. For connected vehicles with lower level of automation that do not have perception sensors, infrastructure sensors will significantly boost its capability to understand the driving context. Even if a full suite of sensors is available on a vehicle with higher level of automation, infrastructure sensors can support overcome the issues of occlusion and limited sensor range. To this end, a cooperative perception modeling framework is proposed in this manuscript. In particular, the modeling focus is placed on a key technical challenge, time delay in the cooperative perception process, which is of vital importance to the synchronization, perception, and localization modules. A constant turn-rate velocity (CTRV) model is firstly developed to estimate the future motion states of a vehicle. A delay compensation and fusion module is presented next, to compensate for the time delay due to the computing time and communication latency. Last but not the least, as the behavior of moving objects (i.e., vehicles, cyclists, and pedestrians) is nonlinear in both position and speed aspects, an unscented Kalman filter (UKF) algorithm is developed to improve object tracking accuracy considering communication time delay between the ego vehicle and infrastructure-based LiDAR sensors. Simulation experiments are performed to test the feasibility and evaluate the performance of the proposed algorithm, which shows satisfactory results.Keywords: cooperative perceptioninfrastructure sensorsobject trackingtime delayunscented Kalman filter Author contributionsThe authors confirm their contribution to the paper as follows: study conception and design: Chenxi Chen, Xianbiao Hu, Zhitong Huang; data collection: Chenxi Chen; analysis and interpretation of results: Chenxi Chen; draft manuscript preparation: Chenxi Chen, Qing Tang, Xianbiao Hu, Zhitong Huang. All authors reviewed the results and approved the final version of the manuscript.Disclosure statementNo potential conflict of interest was reported by the author(s).