Abstract: In response to the problems of low automation levels, high water consumption, and poor effectiveness of dust cleaning in coal mines, a roadway dust cleaning robot is developed. This robot is equipped with functions including automatic monitoring of dust accumulation, autonomous/semi-autonomous movement, and adaptive dust cleaning operation. A vision-based continuous monitoring technology for dust accumulation and cleaning parameters in the spatial domain of tunnels is proposed. An embedded edge AI dust accumulation image sensor is developed based on the HiSilicon HI3519A platform to acquire the spatial distribution characteristics and patterns of roadway dust. The YOLOv8 classification model is employed to train the production dust accumulation heat map, addressing the issues of roadway dust type recognition and dust thickness measurement; SLAM technology integrating multiple sensors such as close-range visible light wide-angle cameras, long-range telephoto infrared cameras, laser lidar, millimeter-wave radar, and combined inertial navigation is adopted, With auxiliary positioning methods including UWB, IMU, and wheel speed sensors, three-dimensional positioning and navigation in roadway sections and directions are achieved. Based on operational processes and safety regulations, the drivable area of the entire roadway section is generated, providing spatial constraints during the autonomous dust cleaning process; A kinematic forward and inverse solution model for the robotic arm is constructed, and a “vehicle arm” collaborative integrated system is developed for the autonomous mobile transport platform of the robot and the collaborative linkage control of the operating robotic arm. A flexible multi-mode adaptive control strategy for the 4-degree-of-freedom robotic arm is proposed. Solving problems such as system collaborative operation planning, high-precision control of end effector pose, and adaptive disturbance rejection control; The combined action mechanism of multiple dust cleaning methods is studied, and an efficient water-saving adaptive dust cleaning strategy is proposed. A full-scene automatic dust cleaning system with adaptive adjustment of high and low pressure spray and rolling brush speed is developed; A prototype of the adaptive roadway dust cleaning robot is manufactured. Ground tests demonstrate that the dust cleaning robot is capable of flushing and wiping the roof, floor, walls, pipelines, and other positions of the roadway; It achieves autonomous obstacle avoidance and “vehicle–arm” cooperative control, with a driving speed of ≥ 2 m/s and a climbing ability of 14°; The flexible obstacle avoidance and explosion-proof robotic arm has a rotation range of 360° and adjustable height of 0−5 m; suitable for roadway section of 16−25 m² and has a continuous operating range of ≥ 1 000 m. The robot plays an important role in reducing the number of underground auxiliary workers and mitigating occupational health hazards for coal mine employees.