Intelligent unmanned agricultural driving training method based on Unity3D and deep reinforcement learning

Abstract

An intelligent unmanned agricultural driving training method based on Unity3D and deep reinforcement learning comprises the following steps: 1,establishing a classic farmland scene in Unity3D, whereinthe classic farmland scene comprises a plurality of environmental parameters; 2, carrying out simulation operation by that driver base on the classic farmland scene, and quantifying the operation process parameters to obtain the operation parameter; 3, constructing a convolutional neural network according to that environment parameter and the operation parameters; 4, pre-training the convolutional neural network to obtain a pre-trained agent; 5, randomly generating a dynamic farmland scene, and adding a random environment parameter into that dynamic farmland scene; 6, putting the pre-trainedagent into a dynamic farmland scene, and carrying out self-training by use a depth reinforcement learning algorithm to obtain a reinforced agent; 7, in the dynamic farmland scene, increasing the frequency of part of the random environmental parameter, and repeatedly training the strengthen agent. The method can rapidly and efficiently improve the driving ability of a driver.

Description

technical field [0001] The invention relates to the technical field of unmanned driving of agricultural machinery, in particular to an intelligent unmanned agricultural driving training method based on Unity3D and deep reinforcement learning. Background technique [0002] With the development of science and technology, agricultural machinery is becoming more and more intelligent, and the automatic driving technology of agricultural machinery is the key technology for efficient agriculture. There are many unknown problems that need to be dealt with in the natural environment where agricultural machinery is located in actual production and life. During operation, agricultural machinery may encounter obstacles such as utility poles, people, and other machinery or approach the ground. At this time, agricultural machinery needs to automatically make obstacle avoidance decisions. Different farmlands require different path planning, so that agricultural machinery can work more effi...

AI Technical Summary

Problems solved by technology

This obstacle avoidance method is too simple to avoid obstacles, and the intelligence level is relatively shallow.

Images