ELM-PSO-GA based inverse solution algorithm for robotic arm

Aiming at the problems of low efficiency of highdimensional nonlinear optimization and easy to fall into local extremes in the inverse kinematic solution of complex robotic arms, The present paper puts forward a hybrid optimization framework (ELM-PSO-GA), which integrates Extreme Learning Machine (ELM), Particle Swarm Optimization (PSO) and Genetic Algorithms (GA). Taking the Mitsubishi RV-4FL-D six-degree-offreedom robotic arm as the research object, its kinematic model is constructed based on the D-H parameter method, and the initial inverse solution is quickly generated by ELM, which is combined with the PSO-GA collaborative optimization mechanism to balance the ability of global exploration and local exploitation, This enhancement has been demonstrated to result in a substantial improvement in the accuracy of the required inverse solution and the convergence speed. The experimental results demonstrate that the mean end error of ELM-PSO-GA is 0.0209 mm, and that the computation time is reduced by $4.3 \%$ compared with the Hybrid algorithm. The algorithm demonstrated its capacity to attain convergence below the error threshold within 30 iterations, a 36% reduction in iterations when compared with the comparison algorithm. This study provides an efficient solution for real-time trajectory planning of high-dimensional nonlinear systems, which effectively guarantees the stability and reliability of the robot arm’s motion under complex working conditions.