In this paper we investigate generating an optimal running gait for the planar model of ATRIAS bipedal robot. ATRIAS is a robotic prototype implemented in Oregon State University with the aim of high speed running. Gait generation for ATRIAS has been done based on SLIP model. Although this passive model is a good base for gait generation, it does not necessarily yield to the best energy efficient solution. So, in this paper via the gradient based method starting from an initial point given by SLIP based control, we search for an optimal pattern for the running gait that minimizes cost of transport (COT) during one complete step. Equations of motion for each continuous time phases, called stance and flight, and models for take-off and touch-down events are derived. Then by parameterization of motors torque profiles using polynomials and solving the direct dynamic model, optimization problem is solved to minimize COT. The optimization is repeated three times by performing the parameterization in terms of polynomials of degrees 3, 4, and 5, to obtain the most efficient torque profiles. The results indicate that for all three shapes of polynomials COT is reduced compared with SLIP based running gait. Moreover, the minimal COT is achieved by torque profiles of degree 4.

Keywords: ATRIAS Biped Robot, Gait Generation, Parameterization, Gradient based Optimization.

Full-Text [PDF 1311 kb]   (4166 Downloads)