A hybrid and adaptive approach to humanoid locomotion : blending rhythmic primitives and feet placement strategies
Year: 2018
Type: dissertation
Abstract: This thesis presents research developed towards adaptable and effective optimization of controllers for humanoid robots’ locomotion. It addresses this goal by combining features of humanoid locomotion and the impact of different environments in its dynamics, along with mathematical optimization and statistical analysis techniques. Controllers for the locomotion of bipedal robots often face challenges regarding their optimization towards different objectives and different environments. We propose an architecture that uses the information gathered in an optimization/exploration phase to adapt to a terrain with partially unknown characteristics. In the exploration phase virtual simulations are used to optimize the parameters of the controller in different terrains. The results of these optimizations are used to identify the unknown terrain characteristics, and these values are used to select the best parameters for this particular environment. The approach was tested in the simulations of an iCub robot on terrains with variable friction, and of the DARwIn-OP robot in ramps with varying slopes. This work brings contributions to the problem of choosing the values for the open parameters of biped locomotion controllers in various situations. Specific issues that are covered relate to: 1) establishing an optimization framework that can be applied to any controller with open parameters, and which results in both safe and well performing behaviors; 2) finding locomotion behaviors that are effective in multiple and distinct environments; 3) adapting the locomotion to environments with varying characteristics, modeled in previous optimizations.
Source: HAL
Author: Pontes, José
Language: English
Subfield: Biomedical Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W4392268504