Research on Autonomous Localization Method for Cross-Arm Installation Robots Based on Multi-Sensor Fusion

To address the positioning challenges of cross-arm installation robots, this paper proposes an autonomous robot localization method based on multi-sensor fusion. First, residual models for vision, inertial measurement unit (IMU), and wheel odometry are derived. A tightly-coupled nonlinear optimization objective function integrating visual-inertial-wheel measurements is then established. The robot pose is obtained by solving this function through sliding window optimization. Experimental results on the EuRoC and KAIST datasets demonstrate that the improved algorithm effectively reduces error accumulation and achieves more robust and precise positioning compared to conventional localization methods.