Abstract: To address localization degradation of coal mine roadway cleaning robots in long, straight roadways where satellite signals are unavailable, difficulties in multi-process coordination, and reliance on manual intervention for attachment replacement, a comprehensive control method integrating autonomous navigation, multi-tool collaborative control, and automatic attachment replacement of fixture is proposed. For autonomous navigation, a fusion localization method combining angle-clustering-based ground segmentation with the Unscented Kalman Filter (UKF) was employed to suppress axial drift, and an improved pose-aware A^* global planning algorithm was designed to eliminate the need for final pose adjustment. For collaborative control, a layered architecture based on a hierarchical finite-state machine was constructed to enable autonomous scheduling of multiple processes. For attachment replacement, a distance-triggered heterogeneous perception fusion strategy with coarse-to-fine staging was proposed; joint-state feedback and visual localization information were continuously weighted and fused to complete the entire attachment replacement process autonomously. An underground industrial test involving a cumulative operating time of 160 h and an advance of 60 m was conducted. Results showed that the coal mine roadway cleaning robot had a mean trajectory error of 0.01 m, a final heading deviation of 3.8°, and a dynamic obstacle avoidance success rate of 92%. The full-process success rate was 80%, and the manipulator end-effector tracking error was 0.12 m. The success rate of automatic attachment replacement was 93.3%, and the mean duration per replacement was 4.8 min. The number of operators per shift was reduced from 5 to 2, a reduction of 60%, while per-capita efficiency increased from 1.8 m³/shift to 2.6 m³/shift.