Network training and unmanned equipment control method and device

Abstract

The invention discloses a network training and unmanned equipment control method and device, and relates to the field of unmanned driving, a training sample is obtained, then, state information corresponding to a target object and state information of acquisition equipment are used as comprehensive state information to be input into a preset attention network, and the attention network is obtained; and obtaining the attention of the acquisition device for the target object. And for each preset driving strategy of the target object, based on the comprehensive state information, determining an evaluation value for the driving strategy, further determining an evaluation value difference degree between the driving strategies in terms of the evaluation value, and taking a deviation between a minimum evaluation value difference degree and an attention degree of the acquisition equipment for the target object as an optimization target. According to the method, the attention degree network is trained, so that the unmanned equipment can select the driving strategy in combination with the attention degree of each target object during automatic driving, and a reasonable driving strategy can be determined when an uncertain number of target objects exist around the unmanned equipment.

Description

technical field [0001] This specification relates to the field of unmanned driving, and in particular to a method and device for network training and control of unmanned equipment. Background technique [0002] In unmanned driving technology, unmanned driving equipment needs to control itself according to the state of surrounding pedestrians, vehicles and other targets during driving, so as to ensure its own driving safety. [0003] In the existing technology, unmanned driving equipment can determine the next driving strategy through the set rules, but the rules need to be set manually, it is difficult to deal with all situations, of course, unmanned equipment can also pass the surrounding The state of the target is input into the neural network to determine the next driving strategy, but usually the input dimension of the neural network is fixed, and the number of surrounding objects is usually not fixed, so it is difficult to judge the driving strategy . [0004] Therefo...

AI Technical Summary

Problems solved by technology

[0003] In the existing technology, unmanned driving equipment can determine the next driving strategy through the set rules, but the rules need to be set manually, it is difficult to deal with all situations, of course, unmanned equipment can also pass the surroundi...

Images