Implementation Method of Deep Learning in the Field of Unmanned Transportation System Collision Avoidance

With the development of highly intelligent society, unmanned vehicles will be widely used. At the same time, the system pressure faced by unmanned vehicles will increase, including the collision problem of unmanned vehicles. How to solve the collision problem and put forward a reasonable scheme to realize the efficient operation of unmanned vehicles in unmanned systems. This paper comprehensively analyzes the characteristics of the problem itself, the choice of the system's running algorithm, the action space of unmanned vehicles, the space state and data feedback of unmanned vehicles, the systematic training mechanism, as well as the instruction sending and data communication in the system space, and obtains the collision avoidance solution of the dual mechanism of interactive deep learning information processing mode and cellular base station information transmission mode. Under this mechanism, deep learning is the main method to process a large amount of hazard data and generate the best instruction. On the basis of powerful computing power, a closed-loop feedback mechanism is established, which collects a large amount of surrounding environment data from scanning, transmits the data to the decision control subsystem in real time through the cellular base station, and the decision control system synthesizes the calculation results to generate control instructions which are transmitted to unmanned vehicles through the base station, so as to avoid collisions.