CICOSim: A Closed-Loop Simulation Platform for Vehicle-Infrastructure Cooperative Algorithms

With the development of intelligent transportation and autonomous driving technologies, cooperation driving autonomous has become a critical path for achieving road safety, traffic efficiency, and enhancing high-level autonomous driving. Simulation platforms for algorithms are vital due to their data generation capabilities, lower costs in the process of research, development, testing, and validation. Simulators in the “Vehicle-Infrastructure Cooperation” domain face following challenges: a lack of high-fidelity, programmable simulation support for integrated “Vehicle-Road-Cloud” cooperative interaction; the unreality of many simulation methods, where algorithms (e.g., decision-making, planning in end-to-end systems) cannot effectively receive environmental feedback in open-loop setups, preventing valid performance statistics; the difficulties to construct corner cases, which cooperative autonomous driving focuses on compared with single-vehicle intelligence. This paper presents a platform designed to address these issues. We propose an architecture designed to create scenarios as close to real environments as possible, featuring the following: firstly, the platform supports the deployment of “Vehicle-Infrastructure Cooperative” autonomous driving algorithms; secondly, to enhance realism and validity, a spatio-temporal coordinate system and a vehicle’s physical-control model for the algorithms are deployed in the simulator; thirdly, CARLA-SUMO co-simulation is introduced to solve the difficulty of constructing traffic flow-induced corner cases.