基于ESKF与图优化的轻量级多传感器融合SLAM系统

针对机器人以及小型UGV车辆多场景运行的需求,同时由于机器人普遍缺乏运算资源,我们设计了一种轻量的、模块化的并且多功能的SLAM建图系统。该SLAM系统可以使用包括3D激光雷达、GPS、IMU等主流传感器,并通过滤波器进行传感器融合,使得系统同时兼顾了较好的鲁棒性与精度。在融合的策略中,针对激光雷达的特点采用了一种滤波器松耦合以及图优化相结合的模块化融合方案,经过实际测试表明,相比于主流的紧耦合方案,该系统可以在精度接近的情况下实现更快的运行速度。此外,该系统可以通过一个3D激光雷达实现同步构建3D点云与2D栅格地图,同时满足了机器人在室内与室外运行的需求。

In order to meet the requirements of multi scene operation of robots and small UGV vehicles,and because robots generally lack computing resources,we design a lightweight,modular and multi-functional slam mapping system.Our slam system can use mainstream sensors such as 3D lidar,GPS and IMU,and fuse sensors through filter,which makes the system have good robustness and accuracy at the same time.In the fusion strategy,according to the characteristics of lidar,we use a modular fusion scheme combining filter loose coupling and graph optimization.The actual test shows that compared with the mainstream tight coupling scheme,our system can achieve faster operation speed with similar accuracy.In addition,our system can simultaneously build 3D point cloud and 2D grid map through a 3D lidar,which can meet the needs of robot operation in indoor and outdoor.