Real-time predictive kinematic evaluation and optimization for biped robots

Abstract

Abstract-Collision-free walking in cluttered environments is still an open issue for humanoids. Most current approaches use heuristics with large safety margins to plan the robot's motion. That way, the chance of collisions can be greatly reduced but the robot movements are limited artificially. In this context, we extend our framework for motion generation and whole-body collision-avoidance by an online predictive kinematic parameter evaluation and optimization: We propose to evaluate the initial parameter set describing the walking pattern by integrating the full kinematic model of the robot. In the model our local optimization technique for collision avoidance is taken into account. Initial parameter sets, which are kinematically infeasible due to kinematic limits or collisions can be identified and adapted before the motion is executed. Additionally, the parameter set is optimized according to a chosen cost function using a gradient method and the step time is adapted according to a desired mean velocity. The optimization method is applicable to different representations of the walking pattern. The method is presented with simulation results obtained with our multi-body simulation. The method is suitable for real-time control, since the optimization can be stopped if it exceeds a predetermined time budget. In that case, an executable but suboptimal result is used. The proposed procedure is executed before each step which makes it very reactive to changes in the environment or in the user input. We have also validated the real-time performance in experiments with our humanoid Lola.

Kinematic optimization for bipedal robots: a framework for real-time collision avoidance

“…The navigation systems have in common that they provide foothold positions, which can be summarized as a set of parameters. This set and, additionally, user chosen parameters or parameters explicitly calculated by the step planner (Hildebrandt et al, 2015;Nishiwaki et al, 2012), as for example torso height or footstep height, describe the robot's desired motion. It separates the handling of geometric constraints imposed by the environment from the motion generation.…”

“…In the following, the proposed methods are introduced with an illustrative example. We review and compare our methods presented in Hildebrandt et al (2016) and Schuetz et al (2014). While we propose in Hildebrandt et al (2016) a parameter optimization to optimize the motion of a bipedal robot, we present in Schuetz et al (2014) a model-predictive approach to exploit the redundancy of an agricultural manipulator.…”

“…We review and compare our methods presented in Hildebrandt et al (2016) and Schuetz et al (2014). While we propose in Hildebrandt et al (2016) a parameter optimization to optimize the motion of a bipedal robot, we present in Schuetz et al (2014) a model-predictive approach to exploit the redundancy of an agricultural manipulator. The illustrative example emphasizes the common characteristics of kinematic optimization of redundant robots and our method's applicability to a wide range of problems.…”

“…with g(q)| te denoting a cost term at t = t e . As described in Hildebrandt et al (2016) the optimization problem can be solved using a gradient method for the parameter p. Fig. 3 shows three snapshots of the resulting task space trajectory.…”

“…First, foot trajectories and the vertical CoM trajectory are planned. These trajectories consist of splines as described in Hildebrandt et al (2016). They depend fully on the parameter p wp .…”