Towards dynamically consistent real-time gait pattern generation for full-size humanoid robots

Abstract

Abstract-We propose a two-stage gait pattern generation scheme for the full-scale humanoid robots, that considers the dynamics of the system throughout the process. The fist stage is responsible for generating semi-dynamically consistent step position and step time information, while the second stage incorporated with multi-body dynamics system is responsible for generation of gait pattern that is feasible and stable on the full-scale multi-degree-of-freedom humanoid robot. The approach allows for very rapid gait pattern regeneration during the swing phase of motion and includes information about present dynamic state when regenerating the new pattern. The paper contains description of a developed method, as well as experimental results proving its effectiveness.

Online regeneration of bipedal walking gait pattern optimizing footstep placement and timing

“…The values are calculated with equation below. The derivation of the equation and more detailed description is presented in [7].…”

“…For safety reasons every time the gait pattern is regenerated we compute more than 3 steps with gait ending in standing position. The footsteps in sagittal plane are calculated in a way to keep the terminal velocity calculated during optimization, while the steps in coronal plane are calculated according to (7). For more details on regular steps calculation refer to [7].…”

“…Since that creates the reference pattern that is discontinuous we interpolate between previous pattern and the new pattern over 150 ms window. More detailed description of method we use to refine the gait pattern with use of MBS can be found in [7].…”

“…2) we take an average locomotion velocity as an input. Based on this we generate the first gait pattern based on a method described in [7]. This stage results in reference of the feet, pelvis and COM trajectories.…”

“…Given the estimated COM position and velocity we use a nonlinear optimization method to find optimized step positions and step timing for the present and proceeding two steps. For calculation of the remaining steps we use previously developed method [7]. The gait pattern regeneration time which includes solving optimization problem and generation of gait pattern with Multi Body System (MBS) in the loop takes less than 40 ms.…”