Globally Consistent 3D LiDAR Mapping With GPU-Accelerated GICP Matching Cost Factors

Article Robotics

Efficient LiDAR Odometry for Autonomous Driving

Xin Zheng et al.

Summary: To address the challenges of traditional tree-based neighbor search and the inefficiency in dealing with ground points parallel to LiDAR beams, a novel efficient LiDAR odometry approach is proposed, utilizing both non-ground spherical range images and bird's-eye-view maps. Additionally, a range adaptive method is introduced to estimate the local surface normal robustly, and a fast and memory-efficient model update scheme is proposed to fuse points and their corresponding normals at different time-stamps. Extensive experiments on the KITTI odometry benchmark and UrbanLoco dataset show promising results, demonstrating the effectiveness of the approach.

IEEE ROBOTICS AND AUTOMATION LETTERS (2021)

Add to Collection

Article Robotics

Unified Multi-Modal Landmark Tracking for Tightly Coupled Lidar-Visual-Inertial Odometry

David Wisth et al.

Summary: The system presented is an efficient multi-sensor odometry system that optimizes visual, lidar, and inertial information in real-time. It utilizes a new method to extract 3D line and planar primitives from lidar point clouds and overcomes the suboptimality of typical frame-to-frame tracking methods. Through passive synchronization of lidar and camera frames, true integration of lidar features with standard visual features and IMU is achieved.

IEEE ROBOTICS AND AUTOMATION LETTERS (2021)

Add to Collection

Article Robotics

BALM: Bundle Adjustment for Lidar Mapping

Zheng Liu et al.

Summary: The letter introduces a local Bundle Adjustment (BA) method for lidar SLAM, which effectively reduces drift and allows large-scale dense plane and edge features to be used. To speed up the optimization process, the method also introduces a novel adaptive voxelization method.

IEEE ROBOTICS AND AUTOMATION LETTERS (2021)

Add to Collection

Article Robotics