Convolution neural network-based lane change intention prediction of surrounding vehicles for ACC

Adaptive cruise control is one of the most widely used vehicle driver assistance systems. However, uncertainty about drivers' lane change maneuvers in surrounding vehicles, such as unexpected cut-in, remains a challenge. We propose a novel adaptive cruise control framework combining convolution neural network (CNN)-based lane-change-intention inference and a predictive controller. We transform real-world driving data, collected on public roads with only standard production sensors, to a simplified bird's-eye view. This enables a CNN-based inference approach with low computational cost and robustness to noisy input. The predicted inference of traffic participants' lane change intention is utilized to improve safety and ride comfort with model predictive control. Simulation results based on driving scene reconstruction demonstrate the superior performance of inference using the proposed CNN-based approach, as well as enhanced safety and ride comfort.