Development of a Monocular Camera-based Relative Distance and Speed Estimation Model for End-to-end Adaptive Cruise Control

In this paper, we propose an end-to-end model that estimates the relative distance and speed between ego and leader vehicles using only a monocular front-facing camera input. Unlike conventional approaches that utilize light detection and ranging (LiDAR) or radar sensors, our camera-based methodology reduces sensor dependency and is cost-effective. Additionally, to reduce model complexity, light weight networks such as the simple bird’s eye view (BEV) scheme and gated recurrent units (GRUs) are applied, enhancing real-time autonomous driving performance. Furthermore, we collect the data using MORAI simulator to train and evaluate the model. Finally, the simulation results validate the feasibility of using the proposed network in the real-time relative distance and speed estimator for autonomous driving systems.