面向退化环境的多传感器融合SLAM方法的研究

Multi-sensor fusion SLAM method for degraded environment

准确的定位和地图信息是实现井下铲运机自主导航行驶的关键。然而,对于井下环境,颠簸和湿滑的路面使编码器里程计提供的信息偏差大,没有显著特征的长巷道即感知退化的环境使雷达等传感器的位姿估计容易产生漂移,获得准确的位置和地图信息在井下环境仍具有挑战性。为实现这一目标,创新地采用松耦合框架和监督算法结合,提出了一种面向退化环境的多传感器融合SLAM(同时定位和建图)方法。此方法使用激光雷达和IMU(惯性测量单元),通过卡尔曼滤波器对IMU的高频数据和激光雷达低速姿态估计进行融合,从而提供更为准确的定位估计。此外,此系统可以自我评估退化环境,当评估到系统进入退化环境时,忽略激光雷达的位姿估计从而减少漂移。通过在井下环境中对此系统进行测试,并与其他主流方法进行对比,证明了系统的鲁棒性和高性能。

Accurate positioning and map information are key to achieving autonomous navigation of underground scraper.However,for the underground environment,bumpy and slippery roads cause large deviations in the information provided by encoder odometers,and long lanes without significant features,that is,perceptually degraded environments,make it easy for sensors such as radar to produce drift in pose estimation.Obtaining accurate position and map information in an underground environment is still challenging.To achieve this goal,this paper innovatively uses a combination of a loosely-coupled framework and supervised algorithm and proposes a multi-sensor fusion SLAM(Simultaneous Localization and Mapping)method for degraded environment.This method uses lidar and IMU(Inertial Measurement Unit)and fuses the high-frequency data of IMU and the low-speed pose estimation of lidar through a Kalman filter to provide more accurate positioning estimation.In addition,this system can self-assess degraded environments.When the system evaluates that it has entered a degraded environment,it ignores the pose estimation of the lidar to reduce drift.In this paper,we tested this system in an underground environment and compared it with other mainstream methods to prove the robustness and high performance of system.