Optimal trajectory generation of serially-linked parallel biped robots

Abstract

In this paper, we propose a new parallel mechanism of biped robots, each of its legs is composed of two 3 - DOF parallel platforms linked serially. The thigh possesses three active linear actuators with a constraint bar and seven passive joints, the shank has two linear actuators to operate pitch and roll motion at the ankle and a rotational actuator for the pitch at the knee. To generate locomotion trajectory for the proposed robot, we adopt the genetic algorithm for minimizing energy consumption. For the fast convergence and efficiency, the RCGA (Real Coded Genetic Algorithms) is used. A trajectory is represented by 4-th order polynomials whose coefficients are chromosomes. Equality conditions reflect the start and end position of swing leg and the repeatability condition for stable walking is used with inequality conditions to describe kinematics constraints for swing motion and ZMP (Zero Moment Point) conditions. To demonstrate the effectiveness and performance of the proposed structure and trajectories, computer simulations were executed using SimMechanics®.