Localization and mapping are crucial components of robotics, especially in the field of autonomous mobile robots. Unmanned forklifts face challenges in self-localization and environmental perception, especially in complex environments with lost sensor signals. This paper focuses on multi-sensor fusion 3D SLAM (Simultaneous Localization and Mapping) using the Error State Kalman Filter (ESKF) to integrate GPS, IMU, and LiDAR data. In outdoor settings, stable GPS signals are combined with IMU to enhance positioning data’s refresh rate and stability using RTK GNSS technology. Indoors, weak GNSS signals make it impractical to resolve fixed solutions, but stable environmental features allow for localization and point cloud map construction using 3D LiDAR and IMU. For indoor-outdoor transitions, dynamic switching of positioning data ensures stability. The forklift is positioned near the outdoor work area, establishing a local coordinate system. GPS pose information is used to initialize the SLAM algorithm for indoor mapping. This fusion algorithm predicts the robot’s posture using IMU data and updates its state with LiDAR and GPS observations, improving mapping and localization accuracy and robustness. Compared to other algorithms, the proposed method shows a 0.05m RMSE, a 54% reduction. Experimental results demonstrate that integrating GPS-IMU-LiDAR enhances positioning and navigation accuracy and refresh rate, suitable for industrial environments with indoor-outdoor routes.