Seamless Positioning and Mapping Using an Adaptive GNSS/INS/LIDAR/Wheel Odometry Integration Based on Factor Graph Optimization

This paper presents LIWO-GO, an extension of the algorithm LIWO-SLAM, which incorporates GNSS in a factor graph for Simultaneous Localization and Mapping (SLAM). To ensure a seamless transitions from outdoor to indoor environments, the GNSS observations must be properly weighted. For this, an adaptive weighting scheme and a trust score are used. The trust score is based on the combination of a hybrid autoencoder for GNSS SNR values with a map-based approach using the SLAM map. To evaluate the algorithm, a tracked robot was equipped with a dual-antenna GNSS receiver, an IMU, and a LiDAR. A test dataset was collected where the robot was steered along a route that contains both outdoor and indoor environments. The trajectory obtained by LIWO-GO is compared to a reference trajectory. The results show that with the trust score and the adaptive weighting scheme, the position estimation of the robot can be improved.