Research On Cooperative Real-Time Position Optimization Method For A Robotic Swarm With Multiple Sensors

With the development of technology and the increasingly complex executable tasks and application scenarios of unmanned devices,collaborative task execution of multiple unmanned devices has become increasingly important.In order to cope with complex unmanned equipment application tasks,real-time accurate positioning and accurate environmental mapping become key issues,namely,SLAM technology.In practical applications,GNSS signals can be obtained to assist unmanned equipment in positioning and mapping.However,most application scenarios,such as underground space,often cannot obtain sustained and stable GNSS signals.Therefore,in order to pursue system efficiency and robustness,multi vehicle SLAM real-time positioning systems equipped with multiple sensors have gradually become a mainstream application solution.However,due to the inherent errors of sensors and accumulated errors in motion,unmanned vehicles equipped with multiple sensors will experience trajectory drift after moving for a period of time.In order to solve the above problems,most SLAM systems are equipped with a loop closure detection module.This article takes the collaborative real-time location optimization method for unmanned vehicle teams equipped with 3D laser radar and IMU as the research topic.Aiming at complex indoor and outdoor loop closure scenarios,under the framework of distributed multi vehicle SLAM system,it further explores the improvement of the loop closure module framework design of multiple unmanned vehicles on their collaborative positioning ability.For the optimization of the loop closure module framework of the multi unmanned vehicle SLAM system,the main work of this article is as follows:1.Lidar-Iris and IKDTree algorithms are proposed to improve the search and matching efficiency of multi-vehicle loop closure descriptors.This thesis uses the lightweight global descriptor Lidar-Iris for scene location recognition.Compared with other descriptors,Lidar Iris contains more effective feature information and robust description information.In addition,this paper will also adopt an incremental search tree algorithm(IKDTree)to accelerate the search process of multi workshop scene descriptor matching;2.This thesis proposes a multi vehicle loop closure detection module with initial relative distance constraints.In the multi-vehicle SLAM system,the initial relative distance relationship of each workshop is a relatively easy to obtain a priori parameter.The addition of the initial relative distance can not only reduce the search range of descriptor matching in the global map building process and the incorrect loop closure that does not meet geometric constraints in similar scenarios,but also further add the initial relative distance to the paired consistent maximum measurement set(PCM)and the improved two-stage distributed pose optimization method,Thus,it can provide a more reliable initial pose guess,speed up the optimization convergence process,and reduce the impact of false loop closure on the optimization process.Comparative experimental results show that the performance of our method is significantly improved compared to the currently most robust distributed multi vehicle Di SCo-SLAM loop closure detection algorithm.