Abstract: At present, the automatic sorting system of gangue mostly adopts series mechanical arm, which can only sort out the gangue with small particle size.The separation speed is slow and the separation effect is not ideal. In view of this problem, a kind of dual-arm parallel coal gangue sorting robot was designed. The robot adopts the sorting mode of pushing rather than grasping, which can sort the gangue with the particle size of 300-600 mm, and greatly reduce the torque demand of the manipulator and improve the sorting speed. In order to avoid rigid impact caused by frequent start-up and stop and prolong the service life of the motor, parabola interpolation function, cubic polynomial interpolation function and quintic polynomial interpolation function are respectively used for trajectory planning in joint space. Simulation and comparative analysis results show that the angular acceleration curve transition is smooth when using quintic polynomial interpolation function for trajectory planning, the angle and angular velocity constraints of the starting and ending points of the robot trajectory are satisfied, and the angular acceleration constraint is also satisfied, so that the motion trajectory of the sorting robot is more stable. The experimental results show that for the coal gangue with particle size of 300-600 mm,the recognition rate and sorting rate of dual-arm parallel coal gangue sorting robot are 91.14% and 86.29% respectively, with high sorting accuracy and stability; it only takes 1.2 s to complete a single complete sorting operation, which greatly shortens the sorting cycle compared with manual sorting; through joint space trajectory planning, the rigid impact of the motor in the sorting process is reduced, which can ensure the system stable work for a long time.