Research On Multi-robot SLAM With Unknown Initial Position Distribution

Study of SLAM technology becomes one of the central issues.The study ofmulti-robot SLAM bases on the study of single-robot SLAM but more valuable thanit.Above all,multi-robot SLAM technology needs to solve single-robot SLAMproblem.The mainly problems in single-robot SLAM mainly include the localization ofrobot-self and landmark,map representation,picking a sensor.The difficulties ofmulti-robot SLAM mainly include the communication between robots and mapsplicing,besides the difficulties in single-robot SLAM.In order to solve these problems,based on the real experiment conditions,the SLAMsystem chooses Kinect camera as robot’s sensor,and uses grid representation and dataassociation localization methods.Data association problem is the most important anddifficult part in SLAM technology.Based on the traditional data associationtechnology,proposed a improved data association method using genetic algorithms andSIFT featuresTo improve the efficiency, robots uses the wireless communication technology basedon UDP protocol and Thread Synchronization to communication.Map splicing problemneed to solve two cores,namely the map align and data association problems.Solve themap align problem by homogeneous coordinate transformation.Experiments showsthat,improved data association algorithm works well on localization.The proposed keymethods can solve the multi-robot SLAM problem well when unknown the robots’ initialposition distribution.