Generation of optimal trajectory for biped robots with knees stretched

Abstract

In previous researches, most of biped robots have walked with their knees bent. However, a human walks without bending his or her knees, which defines “natural” gaits. This is one of the significant differences between human locomotion and biped robot locomotion. This paper proposes a method to generate natural locomotion trajectory without bending knees but based on using a 2-DOF model of linear inverted pendulum mode (LIPM). In addition to upper body motion, this paper proposes foot trajectory that allows heel-lifting and toe-lowering phases like the foot of human. This trajectory is parameterized and optimized with a genetic algorithm. A genetic algorithm is used in order to generate a locomotion trajectory for the proposed robot to minimize energy consumption. The Real Coded Genetic Algorithms (RCGA) is used for fast convergence and accuracy. In simulation, the resulting trajectory spends 20% less energy than the trajectory generated by the LIPM method, which results in bending the knees.