Obstacle Avoidance Method Of Micro-robot Based On Locust Optic Nerve

Micro-robots have been the focus of research in recent years,and have been widely used in swarm robots,environmental exploration,search and rescue and other fields.Obstacle avoidance is an important ability of micro robot autonomous control.Currently,obstacle avoidance relies mainly on infrared,ultrasonic and radar.However,these systems have severe requirements for environmental conditions,so greatly limit the autonomous performance of micro-robots.The vision sensor were applied to the micro mobile robot,aiming at the above limitations,to improve the autonomy of obstacle avoidance.In addition,the hardware and software structure of the vision sensor is complex,and the cost of economy and time is high.Micro-robots have simple structure and limited computing resources.To resolve this contradiction,the physiological research results of locust optic nerve were used for reference,obstacle avoidance neural network in complex environment was constructed,and the obstacle avoidance perception of the micro-robot in complex environment was realized based on this.The main work of this thesis is as follows:(1)CDNN(Collision Detection Neural Network)had poor anti-interference ability to background clutter,a neural network model CDNN-SV(Collision Detection Neural Network with Saturation and Value)for radial motion pattern recognition was constructed by input of color images,based on structural Characteristics of Collision Detection Neural Network,using the HSV color segmentation method,and the corresponding neural network algorithm was proposed.The results show that the computational complexity of the algorithm is determined by the resolution of the input video image.Numerical experiments based on different visual stimuli show that this neural network can not only reveal the specific visual perception characteristics of radial sensitive neurons,but also it can effectively warn collision in complex environment.(2)Micro robots had limited resources and weak autonomy,the hardware and software of the bionic vision system of the micro-autonomous wheeled robot were designed,by adopting the idea from the whole to the details.The complexity of the algorithm and the pressure of the robot's own resources were fully considered,the two were balanced to maximize the autonomous performance of the robot.The results show that the micro-robot uses only one MCU(Microcontroller Unit)for motion and visual control,consuming less energy and cost,compared to the existing visual micro robots.(3)Micro robots had obstacle avoidance problems in real environment,the physical experiment was designed and the operation platform was built.The neural network response and the obstacle avoidance trajectory of the robot were analyzed with the help of the existing equipment,and the control experiment was repeated for 10 times to calculate the success rate of obstacle avoidance.The results show that bionic vision approaches can not only maximize the resource utilization of micro-robots,but also can effectively improve the real-time and autonomy of the micro-robot to avoid obstacles.