Path Navigation Method Of Autonomous Mobile Robot Based On End-to-End Neural Network Model

Autonomous mobile robot is widely used in security patrol,freight,autopilot,and family life and other fields.One of the most important factors to improve the application effect is to be able to accurately navigate the path,so how to design a highly efficient,intelligent autonomous mobile robot navigation algorithms has been a popular topic in this field.However,most of the existing methods for studying the path navigation of autonomous mobile robots are inefficient,costly and have poor robustness and generalization ability.With the increasing complexity of the road environment,the increase of population and buildings,these problems become more prominent.In order to break through the limitations of traditional autonomous mobile robot path navigation methods,this paper proposes an autonomous mobile robot path navigation method based on an end-to-end neural network model,and builds a three-dimensional simulation environment in Unity3 D to automate different models and different environments.The mobile robot path navigation method performs multiple sets of comparative experiments and analysis.The main work and innovation of this paper are as follows:1.Combining the PilotNet model and the Network in Network(NIN)model,a new end-toend neural network model is proposed,named PninNet.The PninNet model reduces the computational complexity of the model's convolution operations by reducing the matrix dimensions and convolution kernel parameters.The multi-channel information integration is realized by the linear combination of multi-feature graphs to improve the expression ability and feature learning performance of the model.At the same time,Dropout is added after each layer of the model to simply and effectively avoid the over-fitting problem in the subsequent training phase of neural network,and further improve the robustness and generalization ability of neural network.2.The PninNet model is used to implement the path navigation method for autonomous mobile robots based on the end-to-end model in the three-dimensional simulation environment constructed in this paper,and an algorithm based on ray detection is used to avoid obstacles.Through a series of comparative experiments(such as comparing with PilotNet model)and analysis,it is proved that the algorithm proposed in this paper improves the efficiency of robot navigation research and the robot navigation performance is better in unknown or familiar environments.3.The method in this paper uses Unity3 D to build a three-dimensional simulation environment to collect training data.On the one hand,it can be collected in a variety of road environments without the support of the experimental base which greatly reduces the time and cost of data collection.On the other hand,it can realize the collection of special data sets(collision,off-road,etc.)that are difficult to carry out due to security risks in the real environment,so as to ensure the integrity of the data,so as to accelerate the convergence speed of the algorithm and enhance the stability of the system.At the same time,the collected data set automatically adds the characteristic label needed by the experiment,which avoids the need of manual marking and greatly improves the experiment efficiency.Finally,combined with the above scheme,the robot model and the environmental road model were established in the three-dimensional simulation environment to conduct multiple simulation experiments,and study and analyze the experimental process and experimental results to improve the defects in the algorithm and experiment.Finally,it is proved that the robot navigation system developed by the method in this paper has a certain generalization ability to unknown environment,target and obstacle,and to a certain extent,it can save the research cost and improve the research efficiency,which has certain reference value for the navigation research of autonomous mobile robot.