Bioinspired Bipedal Locomotion Control for Humanoid Robotics Based on EACO

TL;DR

This paper introduces an enhanced elitist-mutated ant colony optimization (EACO) algorithm for real-time humanoid robot gait synthesis, demonstrating improved convergence, robustness, and practical walking speeds through extensive experiments.

Contribution

The work develops a novel EACO algorithm with genetic and crossover operators, improving real-time optimization for humanoid gait control and addressing stagnation issues.

Findings

Achieved reliable walking gaits with speeds up to 0.47m/s.

Demonstrated improved convergence and reduced stagnation in EACO.

Validated effectiveness across diverse robotics optimization problems.

Abstract

To construct a robot that can walk as efficiently and steadily as humans or other legged animals, we develop an enhanced elitist-mutated ant colony optimization~(EACO) algorithm with genetic and crossover operators in real-time applications to humanoid robotics or other legged robots. This work presents promoting global search capability and convergence rate of the EACO applied to humanoid robots in real-time by estimating the expected convergence rate using Markov chain. Furthermore, we put a special focus on the EACO algorithm on a wide range of problems, from ACO, real-coded GAs, GAs with neural networks~(NNs), particle swarm optimization~(PSO) to complex robotics systems including gait synthesis, dynamic modeling of parameterizable trajectories and gait optimization of humanoid robotics. The experimental results illustrate the capability of this method to discover the premature convergence probability, tackle successfully inherent stagnation, and promote the convergence rate of the EACO-based humanoid robotics systems and demonstrated the applicability and the effectiveness of our strategy for solving sophisticated optimization tasks. We found reliable and fast walking gaits with a velocity of up to 0.47m/s using the EACO optimization strategy. These findings have significant implications for understanding and tackling inherent stagnation and poor convergence rate of the EACO and provide new insight into the genetic architectures and control optimization of humanoid robotics.