Abstract: It uses the sevenlink model for humanoid robots to establish the robot posture vector. First, a polynomical equation with five parameters was used to fit the robot hips curves, and then a nontime reference variable was used to work out the robots ankle and knee trajectories. In the region where the robot can walk stably, the maximum distance for the robots ZMP deviation from the foot center was obtained by using the genetic algorithm. At the same time, the parameters of the hip were determined to optimize the humanoid robot gait. Experimental results show that this method is effective.

Key words: humanoid robots; non-time reference; genetic algorithm; gait optimization