Abstract: As the main coal flow transportation equipment in underground coal mines, belt conveyors may experience coal spillage and accumulation during operation. Traditional manual cleaning methods are inefficient and pose safety hazards. Existing roadway cleaning robots for coal removal suffer from limited operating scenarios and low flexibility of cleaning mechanisms. To address these problems, an automatic coal spillage cleaning mechanism for belt conveyors in underground transportation roadways was designed. The mechanism consisted of a PCC-type coal-raking robotic arm and a CRRR-type excavating and transporting robotic arm, which were carried to the designated position by a crawler-walking platform. The coal-raking robotic arm raked and gathered the coal accumulated beneath the belt, and then the excavating and transporting robotic arm dug up the gathered coal and placed it back onto the belt. Static simulation of the coal-raking robotic arm inserting into the coal pile and horizontal scraping conditions was performed using Ansys Workbench. A transient dynamic analysis of the rack-and-pinion lifting mechanism under frequent start-stop conditions was also conducted. In addition, ADAMS software was used to perform kinematic and dynamic simulations of the autonomous scraping operation of the coal-raking robotic arm and the autonomous excavating operation of the excavating and transporting robotic arm. The simulation results showed that the strength and stiffness of the automatic cleaning mechanism met actual working requirements and had a certain safety margin. Both the coal-raking robotic arm and the excavating and transporting robotic arm operated smoothly to the target position according to the given joint angles, satisfying the coal spillage cleaning requirements under the condition that the narrowest roadway in underground coal transportation was only 80 cm wide. Under varying load conditions, the torque responses of each joint of the cleaning mechanism met expectations, verifying the structural matching with actual task requirements.