Abstract:The central pattern generator (CPG) is widely applied in biped gait generation, and the particle swarm optimization (PSO) algorithm is commonly used to solve optimization problems for CPG network controllers. However, the canonical PSO algorithms fail to balance exploration and exploitation, resulting in reduced optimization accuracy and stability, decreasing the control effectiveness of CPG controllers. In order to address this issue, a balanced PSO (BPSO) algorithm is proposed, which achieves better performance by balancing the algorithm's exploration and exploitation capabilities. The BPSO algorithm's solving process consists of two phases: the free exploration phase (FEP), which emphasizes exploration, and the attention exploration phase (AEP), which emphasizes exploitation. The proportion of each phase during optimization is controlled by an adjustable parameter. The BPSO algorithm is subjected to qualitative, numerical, convergence, and statistical analyses based on 13 benchmark functions. The experimental results from the benchmark functions demonstrate that the BPSO algorithm outperforms other comparison algorithms. Finally, a linear walking optimization method for humanoid robots based on the BPSO algorithm is established and tested in the Webots simulator. Comparative results with two other optimization methods show that the BPSO‐based optimization method enables the robot to achieve greater walking distance and smaller lateral deviation within a fixed number of iterations. Compared to the other two methods, walking distance increases by at least 60.98% and lateral deviation decreases by at least 1.96%. This research contributes to enhancing the locomotion capabilities of CPG‐controlled humanoid robots, enriching biped gait optimization theory and promoting the application of CPG gait control methods in humanoid robots.

Biped Gait Optimization Using Estimation of Distribution Algorithm.

Biped Gait Optimization Using Spline Function Based Probability Model.

Estimating Probability Distribution with Q-learning for Biped Gait Generation and Optimization.

Estimating Biped Gait Using Spline-Based Probability Distribution Function with Q-Learning.

Dynamic Walking Gait Designing for Biped Robot Based on Particle Swarm Optimization

Optimal On-Line Walking Pattern Generation for Biped Robots

Time-Variant Gait Planning for Under-Actuated Biped Robot Via Optimization

Stable Biped Walking Based on Linear Search Optimization Algorithm

Gait Optimization and Energy-Based Stability for Biped Locomotion Using Large-Scale Programming

Gait Optimization for Legged Systems Through Mixed Distribution Cross-Entropy Optimization

Distal Learning Applied to Biped Robots

3D walking planning optimization for a biped robot using genetic algorithm

Energy-efficient Bio-Inspired Gait Planning and Control for Biped Robot Based on Human Locomotion Analysis

Optimal Gait Design for a Soft Quadruped Robot via Multi-fidelity Bayesian Optimization

Optimization of Biped Robot Walking Based on the Improved Particle Swarm Algorithm

of Advanced Robotic Systems Real-time Walking Pattern Generation for a Biped Robot with a Hybrid CPG-ZMP Algorithm immediate

Gait Optimization Method for Humanoid Robots Based on Parallel Comprehensive Learning Particle Swarm Optimizer Algorithm

BipedWalking Control Using Offline and Online Optimization

Gait Optimization of A Novel Tripedal Robot

Optimal Walking Gait Generator for Biped Robot Using Modified Jaya Optimization Technique

Research on Walking Gait Algorithm for Biped Soccer Robot