Research On Path Planning Of Mobile Robot In Dynamic Scene

With the development of the times and the progress of science and technology,the visual SLAM system has been used in many fields,such as automatic driving,logistics storage robots,etc.However,the relatively mature SLAM systems in the past are mostly based on the assumption of static scenes,that is,objects in the current mapping and navigation environment are static.This makes the SLAM system unable to be applied in a more universal environment,because there will inevitably be dynamic objects in real life,such as storage robot transportation.In this environment,there will inevitably be multiple robots working together.The previous multi robot storage system uses centralized,that is,the host is responsible for task allocation,so that the robot’s running track does not conflict.However,in this way,First,it does not conform to the characteristics of distributed tasks.Second,if the host is disconnected,the entire robot system cannot run.In order to avoid this situation,this paper proposes a SLAM system for dynamic scenes.The system mainly includes real-time dynamic/static region extraction based on ORB algorithm,dynamic object trajectory prediction system based on adaptive fractional optical flow,path planning system with added virtual obstacles,finally,converted into chassis motion control of the robot.The real-time region segmentation algorithm based on ORB algorithm is an improvement on the previous ORB feature algorithm.It distinguishes the dynamic/static objects in the screen by calculating the relative moving distance of the current frame,and performs mask segmentation.At the same time,feature matching and sparse point cloud mapping are carried out for static objects.The fractional order optical flow system is a system based on LK sparse optical flow.In the past,the optical flow system was first order.If the order is fixed,the image information will be lost or the image details will be too much,leading to false recognition.In order to solve this problem,an adaptive fractional order optical flow system is proposed.Based on the image signal-to-noise ratio,different fractional orders are assigned to images.For example,in the case of under exposure,the fractional order is reduced to increase image details.After verification,the system improves the robustness of the optical flow system,reduces the impact of photometric changes on the system within a certain range,and increases the image characteristics when smoothing the gradient.The local trajectory planning algorithm is mainly based on the artificial potential field method.The local artificial potential field method is used to ensure the real-time path planning.A method based on vector transformation is proposed to avoid the artificial potential field trajectory planning entering the dead zone.Of course,this paper also adds the dynamic tracking of the optical flow system to the path planning in the form of virtual obstacles,so that the mobile robot carrying the SLAM system has a certain obstacle avoidance ability.