ALIW-IESEKF: Tightly coupled wheel-LiDAR-IMU SLAM for unstructured agricultural environments

To address the challenges of feature sparsity, geometric degradation, and motion distortion in unstructured agricultural environments, this paper proposes ALIW-IESEKF, a robust Wheel-LiDAR-IMU tightly-coupled Iterated Error State Extended Kalman Filter odometry algorithm. The core contributions are twofold. First, we derive a full-dimensional observation update equation that incorporates both linear and angular velocities from wheel odometry as hard constraints within the IESEKF framework, effectively reducing scale drift in LiDAR-only odometry under geometric degradation. Second, we integrate an adaptive voxel filtering strategy and a hash-octree-based nearest neighbor search mechanism, significantly enhancing map maintenance efficiency and query precision in complex agricultural settings.Experimental results on the public CitrusFarm dataset and real-world agricultural scenarios demonstrate that ALIW-IESEKF reduces the Absolute Trajectory Error (ATE RMSE) by 13.7% to 47.7% compared to state-of-the-art algorithms such as FAST-LIO2 and RSS-LIWOM. Quantitative analysis demonstrates that our system achieves superior map consistency with 2.694 bits average entropy (13.6% reduction from FAST-LIO2) in GPS-denied betel nut forests, while qualitative visual inspection confirms sharp boundary reconstruction and ghosting artifact mitigation in open fields, highlighting the framework’s significant practical value for autonomous agricultural robots.