Particle Swarm Optimization aided PID gait controller design for a humanoid robot

Gait planning for the humanoid robot is a very essential and basic requirement. The humanoid robot is balanced at two feet; therefore, special attention is required for gait analysis for the execution of assigned tasks. In this paper, the linear inverted pendulum (LIPM) model is considered to simplify the study and to obtain better gait planning of humanoid robot NAO. Center of mass (COM) and zero moment point (ZMP) criterion are applied with the LIPM model for a better understanding of selecting the step length and period. In addition, a PSO (particle swarm optimization) tuned PID (proportional-integral-derivative) controller has been implemented. Sensory data such as the location of obstacles and the target along with the desired trajectory aided inverse kinematics have been embedded to the conventional PID controller, which provides an interim angle to start the navigation. This interim angle has been carried forward to the PSO technique accompanied by the desired trajectory. It tunes the parameters of the conventional PID controller and provides an optimum turning angle, which avoids obstacles and increases the stabilization of the robot while crossing it. It reduces travel time and shortens travel length. PSO technique minimizes the computational complexity and number of iteration because it requires fewer tuning parameters. Simulations are executed on the simulated NAO robot for the conventional PID controller and the proposed controller. To ratify its findings, experiments are carried out on a real NAO robot in laboratory conditions for both the conventional PID controller and the proposed controller. Simulation and experimental results are presenting a good agreement among each other with deviation under 6%. Applying the PSO tuned PID controller provides a predictable gait and reduces the stabilization time and essentially eliminating the overshoot by 25%. A comparative study with various controllers is performed, and the credibility of the evaluated result has been examined using statistical analysis. The proposed controller has been compared with a previously developed technique to ensure its robustness.