Automated Evolutionary Gait Tuning for Humanoid Robots Using Inverse Kinematics and Genetic Algorithms

Humanoid bipedal walking remains challenging due to unstable, high-dimensional dynamics and the labor-intensive, platform-specific tuning typically required to obtain workable gaits. We present a hybrid framework that couples a compact screw-theory kinematic model with a multi-objective genetic algorithm (GA) to tune humanoid gait parameters automatically. The method parameterizes the foot’s half-elliptical swing (horizontal and vertical speeds) and the torso pitch angle, and optimizes stride length while limiting lateral deviation through a single, weighted objective. Relying only on kinematic models—without explicit dynamic equations—the framework integrates inverse kinematics and Jacobian computation to evaluate candidate solutions efficiently. We validate the approach in simulation and on a 14-degrees-of-freedom (DoF) humanoid platform. This work contributes a compact modeling and optimization strategy that enables sim-to-real transfer, establishing a foundation for future extensions incorporating stability criteria, sensor feedback, and adaptive weighting.