Applying the Divergent Component of Motion Method for Quadrupedal Locomotion to a Robot with Series Elastics Actuators

Abstract

Many approaches for the trajectory generation for dynamical locomotion were developed in order to handle the hybrid and non-linear nature of locomotion of compliant four-legged robots. However, applied methods such as Central Pattern Generator, Finite State Machine, model-predictive control or non-linear optimization approaches do not allow a direct and analytical generation of smoothed CoM trajectories. The Divergent Component of Motion method is a trajectory generation method which is already successfully applied to bipedal robots. With this method the direct generation of continuous CoM trajectories of quadrupedals could provide advanced capabilities for the control of dynamical locomotion in challenging terrain. The aim of this thesis is to apply the DCM method to a compliant quadrupedal robot in order to generate CoM and limb trajectories which are executable for this robot. For this purpose a sequnece of a compact stance description is used to describe the gait pattern of trotting, pacing and walk without any flight phases. Based on the stance description the geometric center point of the support polygon is calculated for a heuristic VRP placement rule. This placement determines the VRP waypoints for the robot Bert equipped with planar legs. These waypoints are used with the efficient DCM-method to generate the CoM references. Further, these references are used in combination with a cubic interpolation of the limb movement to generate the joint references by a resolved motion rate approach. The implemented DCM-based trajectory generation was tested in a simulation framework for the trotting and walking Bert. The joints were controlled by a PD-controller during the simulation. In conclusion, the multi-body simulation results show that DCM can be used to generate executable references for four-legged compliant robots. However, these results still need to be validated by experiments on the robot Bert.