Research On Collision Prediction Of Robot Motion Control Based On Axis Data

The background product is X-ray Digital subtraction angiography device,which is a product based on X-ray exposure,through real-time imaging,interventional surgery on blood vessels.It is mainly used in cardiovascular and cerebrovascular diseases,lung cancer treatment embolization,ECG defibrillation,etc.It is the most important medical device for vascular interventional therapy.The core function is the real-time control system,like human brain to monitor and schedule all the sub-functions co-work under predefined rules.Anti-collision was much concerned by the doctor who operates the machine during surgery.Anti-collision is used to protect the operator in safe environment which help them have better user experience.The traditional method is implemented by the prepared 3D models as static data calculated through the kinematic logical in real-time during operating the machine.The computational complexity and high loading limits the development of this technology.This thesis proposed one solution to build cluster math model learning the collision status by itself instead of the complicated transformation and busy geometry calculation.The cluster is implemented by machine learning through the axle joints data and relative parameters.Considering the movement application for the escape strategy problem,this thesis also provides one method to cover this requirement.According to the collision characteristics of machine movement,a new prediction model is designed,which changes the classification problem into a logic regression problem,and realizes the collision escape strategy by predicting the shortest distance.For the cluster design,the following experiment also gives analysis on the accuracy and performance compared the BP neural network and KNN clustering.According to the analysis of the practical application problem,like different axle adjustment parameters,axle movement path coverage,and different classify method of distance range,the cluster needs to be optimized and debug based on experiment results.For the design of collision escape prediction,the paper optimizes the initial weight and threshold of neural network through genetic algorithm to carry out comparative experiments,and finds out the optimal network design which is suitable for practical application.The experimental result shows that the proposed method greatly reduces the operation time of collision state,and the method simplifies the system engineering and improves the stability.This solution also cuts the 3D models mesh work which help to save the project cost on human resource.On the long term,the involved machine learning technology has brought the possibility of unlimited development space for robot technology applications and further advanced human medical device technology.