Optimized Joint Trajectory Model with Customized Genetic Algorithm for Biped Robot Walk

Abstract

Biped robot locomotion is one of the active research areas in robotics. In this area, real-time stable walking with proper speed is one of the main challenges that needs to be overcome. Central Pattern Generators (CPG) as one of the biological gait generation models, can produce complex nonlinear oscillation as a pattern for walking. In this paper, we propose a model for a biped robot joint trajectory in order to be able to walk straight, exploiting polynomial equations for the support leg's joints and Truncated Fourier (TFS) Series equations for the swing leg's joints in the sagittal plane and frontal plane. Four customized genetic algorithms (GA-1 to GA-4) with different implementations for the crossover steps are used as evolutionary algorithms to optimize equation parameters and achieve the best speed and performance in walking motion. These four GAs differ in crossover step and parent selection parts. After a primary evaluation to make sure the next generation is better off than before, we consider a clever comparison feature between the best of two generations (parent and child) in GA-4. The algorithms have been tested on the Darwin humanoid robot in the Webots simulator environment where the results show that the GA-4 model has the best performance and achieves the desired fitness value.