Robust Localization of Autonomous Excavators via EKF-Based Sensor Fusion in Real-World Construction Environments

Abstract

Reliable localization is a prerequisite for autonomous excavation, yet outdoor construction environments introduce GNSS degradation, vibration, and terrain-induced slip that can compromise single-sensor estimates. This study reports the implementation and field evaluation of an EKF-based localization system that fuses Global Navigation Satellite System–Real-Time Kinematics (GNSS-RTK), inertial measurement unit (IMU), and encoder measurements within a robot operating system (ROS) pipeline for an autonomous excavator. The fusion strategy weights complementary sensor measurements according to modeled uncertainty and down-weights GNSS updates classified as degraded. Experimental evaluation on a mapped outdoor route shows that the fused trajectory remains stable under non-ideal operating conditions, including signal obstructions and high-vibration segments. The EKF-fused trajectory achieved a path-deviation RMSE of 0.139 m and a P95 path deviation of 0.213 m relative to the independently surveyed reference path. These results support the feasibility of probabilistic multi-sensor fusion for autonomous excavation tasks in challenging field environments.