Research Progress on Autonomous Driving of Automobiles Based on Deep Learning

Abstract

With the wide application of deep learning in fields such as image processing and computer vision, research on autonomous driving based on deep learning has become a research hotspot in autonomous driving technology. In deep learning, the application of Convolutional Neural Network models can significantly enhance the performance of autonomous driving in environmental perception. Combined with multi-sensor fusion technology, it has considerably improved the accuracy of target detection and semantic segmentation. Based on the information output from the perception layer, at the decision-making and control level, the Recurrent Neural Network model and the Transformer model further play a role in optimizing path decision-making and vehicle control strategies. Finally, through an end-to-end algorithmic framework, imitation learning and reinforcement learning are integrated to achieve the synergy of perception, decision-making and control, thereby significantly enhancing the overall performance of the autonomous driving system. This article will conduct a review of the application of the above-mentioned models and point out the challenges faced by deep learning based on the actual application scenarios. On this basis, it concludes that future research on autonomous driving based on deep learning will develop in the directions of model lightweighting, model refinement, and multi-technology integration.