Normalized Neural Network for Energy Efficient Bipedal Walking Using Nonlinear Inverted Pendulum Model

Abstract

In this paper, we present a novel approach for bipedal walking pattern generation. The proposed method is designed based on 2D inverted pendulum model. All control variables are optimized for an energy efficient gait. To obviate the need of solving non-linear dynamics on-line, a deep neural network is adopted for fast non-linear mapping from desired states to control variables. Normalized dimensionless data is generated to train the neural network, therefore, the trained neural network can be applied to bipedal robots of any size, without any specific modification. The proposed method is later verified through numerical simulations. Simulation results demonstrated that the proposed approach can generate feasible walking motions, and regulate robot's walking velocity successfully. Its disturbance rejection capability was also validated.

An overview on bipedal gait control methods

“…For the locomotion control of biped robots, Wang et al [91] applied a normalised NN to execute the fast solving of non-linear dynamics via the inverted pendulum model, and finally achieved the energy-efficient gait. Castillo et al [92] provided an adaptive NN-based approach to adjust the walking velocity of the biped robots.…”

Modeling and design of an observer-based robust controller for a low-cost inverted pendulum based on the H∞ approach

“…Some studies implement neural networks for various tasks, like in [1] where they use this approach to ease the tracking problem for inverted pendulum systems regarding the repeating tasks of calculating feedforward friction compensations. Similarly, in [2] a neural network is trained to prevent the need to solve the nonlinear dynamics of a bipedal walking system from an inverted pendulum model.…”

An Upper Limb Fall Impediment Strategy for Humanoid Robots

“…This shortcoming is one of the key limitations that restrict the application in more general workspaces of these robots. To overcome this problem, there has been various balance control methods proposed, which can be categorized into: ankle [14], hip [23], stepping [15,19], heuristic [5] and online learning [21] strategies. However, if the external force exceeds the maximum capability or an obstacle constrains the stepping space, there is still a risk of falling.…”