Design Of Desert Hexapod Bionic Robot Based On Deep Learning

Desertification destroys the balance of the ecosystem.It is necessary to study the causes of desertification and to control desertification.However,due to the complex terrain of the desert and the harsh climate,the existing technical equipment is limited by the environment.Fast-moving robots in the environment have become very urgent.This article combines the advantages of the wheeled robot and the footed robot to develop a crescent-shaped six-legged bionic robot.At the same time,a deep learning framework is applied to the embedded device,so that the robot can better adapt to the desert environment.First,the design of the robot's leg structure,transmission device,and overall layout is completed.The leg adopts a new type of crescent wheel structure,and the hexapod coordination can imitate the walking gait of the hexapod insect.The concept of "following angle" is introduced to solve the problem of robot sinking on soft sand foundation,and a control strategy of "variable following angle" is given.The size of "following angle" can be adjusted according to the sinking depth of the leg wheel on the sand foundation,so that It passed through the soft sand foundation smoothly,and explained in detail how to realize the gait and turning gait of the tripod.Secondly,Jetson Nano is the master control chip,STM32 is the slave control chip,the servo motor is used as the actuator to drive the leg joints,combined with the driver and various sensor modules,the hardware circuit and communication network of the desert hexapod bionic robot are constructed to form the The hardware part of the robot control system.On the basis of completing the hardware design of the control system,to achieve data connection at the software level,part of the program design takes Jetson Nano as the core,which is mainly responsible for the processing of environmental information,and the other part of the program uses STM32 as the core,which is responsible for the coordinated control of the legs.After receiving the data from Jetson Nano,control the motor driver.Using the deep learning framework Pytorch,a desert environment recognition method based on AlexNet and transfer learning is proposedCombined with the structural characteristics of the crescent wheel legs,the desert hexapod bionic robot can complete the exploration,rescue and military reconnaissance of complex terrains such as deserts,grasslands and mountains.This robot has a certain reference value for the development of mobile robots for unstructured terrain such as deserts.