Abstract: Aiming at the low working efficiency of medium and heavy duty mechanical arms caused by large joint impact and long running time in engineering applications, a trajectory optimization method of mechanical arms based on the variation strategy particle swarm optimization algorithm was proposed based on inverse kinematics theory and polynomial interpolation. Population variation was introduced into the iteration of the basic particle swarm optimization algorithm and parameters such as inertia weight were adjusted adaptively. Improve algorithm convergence speed and accuracy. In this paper, the minimum running time is used as fitness function to study the changes of angular velocity and angular acceleration of 6 joints of the PUMA560 manipulator under the control of the "5-5-5" joint space trajectory model under different position conditions. Based on the mutation strategy particle swarm optimization, the minimum running time of the manipulator is studied in the range of −π-π rad/s under different position conditions. Matlab was used as the simulation platform to simulate the motion trajectory of the manipulator under two-point cycle and three-point noncollinear conditions and to study the optimal fitness value, the number of iterations when converging, joint position and pose curve, etc. The experimental results show that the two-point cycle operation time is reduced by 50.01% and the three-point non-collinear time is reduced by 67.83% compared with the initial planning time. Compared with the basic particle swarm optimization algorithm and adaptive particle swarm optimization algorithm, the optimal fitness value decreases and the convergence speed increases by 48.28% and 67.77%, respectively. Under the two conditions, the position and pose curves of 6 joints were continuous, without abrupt change, and the running time was shortened.