Reseach On Improved Optical Flow Based Mobile Robot Navigation Method

As mobile robots are more and more widely used in military,industrial,human life and many other fields,the robot working environment is changing from simple and structured environment to flexure unstructured environment such as field,planetary et al.How to effectively obtain the various information of the object in the unstructured environment,so as to carry out the robot's autonomous navigation is an important topic in the fields.The visual sensor has the more advantages than other sensors.What? more,vision is the primary means of robot to get information,just like human and other animal.With the researchers into the study of flying insects? visual nervous system,it is found that flying insects use compound eyes by detecting the time domain changes of visual images(i.e.optical flow)to control their own takeoff,landing,obstacle avoidance and so on.Inspired by the mechanism of insect flight,the obstacle avoidance algorithms based on optical flow has been paid more and more attention.At present,the way of optical flow based obstacle avoidance is relatively simple,usually by balancing the left and right sides of the optical flow magnitude.In the actual complex scenes,this method will lead to frequent changes in the direction of the robot,and for the symmetry of the scene it cannot achieve obstacle avoidance.So,in this paper the robot navigation method based on optical flow is studied,for the optimization of navigation method,the following three aspects were included:(1)To improve the traditional optical flow algorithm,which is sensitive to light change,poor noise immunity and not suitable for large displacement motion,a improved algorithm is proposed.Using the image sequences and real scene sequences,the improved optical flow algorithm is compared with two excellent optical flow algorithms TV-L1 and SimpleFlow.The experimental results show that the proposed method can effectively deal with the changes of light,noise,occlusion and discontinuous regions,improve the accuracy,performance is better than the other two algorithms.(2)A mobile robot obstacle avoidance algorithm based on the improved optical flow algorithm is proposed.The algorithm uses the continuous image sequence taken by the monocular camera as the input,and the time to contact is calculated by the optical divergence.Then use the average of TTC to judge whether to need to avoid obstacles.If needed,through the design of the method to get the angle,control the robot to avoid obstacle.The experimental results show that the algorithm can realize the mobile robot walking without collision in the unknown environment.(3)Since the obstacle avoidance algorithm based on optical flow is aimlessly free to walk and cannot realize the robot navigation.The relative depth TTC calculated by the optical flow is used to construct the repulsive potential field of the artificial potential field.So Artificial field method and optical flow method are combined to control robot to avoid obstacles and reach the target.The experiment proves the effectiveness and feasibility of this algorithm.The subject is supported by the National Science Foundation(NO.61573029)and Beijing Natural Science Foundation Projects(NO.4162012).The research of subject has certain reference value for the study of biomimetic mobile robot navigation.