Research On Indoor Environment Perception And Navigation Algorithm Of Quadruped Guide Dog

More than 295 million people around the world have impaired vision.Visual impairment affects their physical and mental health and quality of life,resulting in reduced mobility.50 %of blind people have never traveled independently,and 40 % of blind people travel less than once a week on unfamiliar routes.Hand-held guide sticks and wearable guide devices have the advantages of simple structure,economy and portability,but they also have the disadvantage of limited sensor capacity.The wheeled guide robot has a more perfect perception system.It is helpful to improve the safety of the visually impaired people through guided walking.However,due to structural limitations,the terrain adaptability and flexibility are insufficient.The quadruped guide machine dog is designed by bionic canine animals.Compared with the above three,it has better environmental adaptability and maneuverability,and is an intelligent guide means with more prospects and research value.In the current indoor complex scene,there are some problems such as positioning drift,incomplete obstacle detection and low efficiency of path planning when the quadruped robot dog is equipped with 2D Li DAR and depth camera for environmental perception and autonomous navigation.To this end,this paper has carried out targeted research,completed the state estimation,environment perception and navigation algorithm of quadruped guide robot dog based on multi-sensor data fusion,and realized the accurate positioning,map construction and autonomous navigation of quadruped guide robot dog in indoor environment.The main contents of this paper are as follows:Firstly,according to the requirements of the quadruped guide robot dog for positioning,perception and navigation capabilities,an environment-aware positioning system combining encoder,Inertial Measurement Unit(IMU),2D Li DAR and depth camera is designed.For the designed sensing positioning system,sensor selection,installation,single sensor internal parameter calibration and multi-sensor external parameter joint calibration are carried out;Secondly,a quadruped robot dog pose and state estimation algorithm based on Invariant Expanded Kalman Filter(IEKF)was proposed,which solves the problem of divergent filtering results and inconsistent observation and state estimation when noise is introduced in the Extended Kalman Filter(EKF)algorithm,and has high frequency and robustness;Thirdly,an environmental perception algorithm combining point cloud fusion and map fusion based on 2D Li DAR,depth camera,and odometer data has been proposed,which solves the problem of map size distortion caused by Li DAR point cloud motion distortion during Synchronous Localization and Mapping(SLAM)in complex indoor environments,as well as the problem of incomplete information when using a single sensor for obstacle detection;Fourthly,an improved A* algorithm for optimizing hop search strategy was proposed,which was combined with the DWA algorithm and voice interaction system to achieve dynamic obstacle avoidance and integrated navigation.Through the state estimation algorithm,multi-sensor fusion environment perception and map construction algorithm and intelligent voice navigation algorithm proposed and designed above,the high-precision SLAM and real-time obstacle avoidance navigation of quadruped guide robot dog in indoor environment are realized.