Research On Human-Robot Interaction Navigation And Path Planning For Indoor Mobile Robot

Due to the unmanned and intelligent characteristics of mobile robots,they play an important role during the new coronavirus epidemic,and have been widely used in the world.Considering the complex application scenarios and diverse tasks of mobile robots,how to flexibly interact with humans for navigation and efficient path planning is a difficult problem faced by the popularization and application of mobile robots,and it is also a research hotspot in the current academic and industrial circles.Therefore,in view of the above problems,this paper studies the human-computer interactive navigation and path planning algorithm of indoor mobile robots.The main contents are as follows:Firstly,aiming at the influence of sensor error on navigation and positioning accuracy during the operation of mobile robots,the modeling and data preprocessing of sensors carried by mobile robots are introduced.This part is an important part of ensuring accurate mapping and accurate positioning and navigation of robots.As the main ranging sensor,single-line lidar needs to remove the laser distortion caused by robot motion in order to ensure its ranging accuracy.The wheel odometer is mainly used to calculate the trajectory of the robot,and the internal parameter calibration of the wheel is an important part of it.The camera mainly plays the role of target detection,and obtaining accurate target images can improve the accuracy of object recognition.The inertial measurement unit is mainly used to assist the pose estimation of the robot due to its high-frequency attitude measurement characteristics.Secondly,different navigation control strategies are designed for the dense crowd interaction and sparse crowd interaction problems that mobile robots often face in indoor scenes.Aiming at the human-computer interaction scene of dense crowd,this paper proposes a deep reinforcement learning navigation algorithm based on multi-attention mechanism fusion,which integrates the probability grid map of the robot to the surrounding crowd,and encodes the perceptual information of the historical time series through the Transformer algorithm to predict the best speed control instruction of the robot in the future.For the human-computer interaction scene of sparse crowd,based on the traditional dynamic window algorithm,the social force model is integrated to fully express the influence of pedestrian state on the robot’s motion,making it more in line with the human-computer interaction navigation scene.In order to test the effectiveness of the above two algorithms in the corresponding scenarios,this paper designs simulation human-computer interaction experiments in dense scenes and sparse scenes respectively.Through the analysis of navigation time,path length and other performance indicators,the algorithm can effectively solve the problem of human-computer interaction navigation in dense and sparse scenes.Finally,aiming at the problem of low efficiency of path planning in mobile robot search task with unclear target position,a path planning algorithm based on belief decision criterion is proposed.In this paper,the prior knowledge of the environment and tasks is innovatively added to the path planning decision-making process of the robot,and a confidence decision criterion with path length and orientation information is developed.Prior knowledge is mainly reflected in the modeling and updating of the confidence state of each room in the target and the environment,as well as the spatio-temporal modeling of the traversable mode between rooms in the environment.The update of the confidence state is mainly affected by the semantic information of the object and the room,and the space-time traversability between the rooms is mainly affected by human activities.In order to improve the accuracy of target detection and ensure the smooth progress of the handling work,a robot posture fine-tuning strategy is proposed to make the target detection and positioning more accurate.In this paper,simulation experiments and physical experiments are designed to verify the effectiveness of the algorithm.Through the analysis and comparison of performance indicators such as task execution time and navigation path length,it can be seen that the algorithm can effectively solve the problem of low path planning efficiency of mobile robots in search tasks with unclear target location compared with greedy algorithm,region first algorithm and uniform algorithm.