Design insights and comparative evaluation of a hardware-based cooperative perception architecture for lane change prediction

Research on lane change prediction has gained attention in the last few years. Most existing works in this area have been conducted in simulation environments or with pre-recorded datasets, and they often rely on simplified assumptions about sensing, communication, and traffic behavior that do not always hold in practice. Real-world deployments of lane-change prediction systems are relatively rare, and when they are reported, the practical challenges, limitations, and lessons learned are often under-documented. This study explores cooperative lane-change prediction through a real hardware deployment in mixed traffic and shares the insights that emerged during implementation and testing. The studied architecture integrates stereo-camera perception, wireless communication, a knowledge-graph-based intention prediction module, and automated longitudinal control implemented on embedded platforms. It is implemented on an ego vehicle and a target vehicle and evaluated in a three-vehicle scenario where a third vehicle acts as the preceding vehicle that forces the target vehicle to change lane. Real-road experiments show that, when the cooperative prediction module is enabled, the ego vehicle can anticipate the target vehicle’s lane-change intention about 4 s before the actual lane crossing and decelerate early to open a safe gap, whereas disabling prediction leads to late reactions and aggressive braking. The experiments also reveal constraints that are critical for real deployments. Perception pipelines are sensitive to outdoor lighting, so tests must be scheduled at times and locations with more stable illumination. A precomputed lookup table keeps prediction fast on embedded devices. Communication reliability and thermal effects on the hardware, especially in hot weather, can noticeably affect the system behavior. By documenting these experiences together with the observed behavior of the vehicles with and without prediction, the study provides practical guidance for others working on similar cooperative prediction systems.