A Control Decision Model for Automatic Driving Based on Monocular Depth Estimation

As one of the key tasks in automatic driving, vision-based depth estimation plays an important role in environment perception and decision control. In this study, a novel control decision model based on monocular depth estimation is proposed for automatic driving. The proposed scheme develops an improved deep learning structure to establish the backbone network of monocular depth estimation method. This is expected to improve the accuracy and robustness of depth estimation. To verify the validity of the proposed model, a large number of experiments and analyses were carried out in this study. In the experiment, several common automatic driving control decision models such as Semantic depth, Fast depth, ResNet50 and Bevgpt are selected as comparison objects, and the performance of the proposed models is comprehensively evaluated through the accuracy evaluation under different epochs. The experimental results show that compared with the comparison model, the proposed model exhibits higher accuracy and stability in depth estimation tasks. This research result provides new ideas and methods for the development of autonomous driving technology, and is expected to make important contributions to the realization of intelligent transportation systems.