MPC-based Minimum Time Humanoid Robot Footstep Planning

Robots are becoming increasingly common across various domains, performing tasks with little to no supervision. Among them, humanoid robots are best suited for human environments, and are likely to eventually become part of daily life. However, precise movement is essential for effective interaction, as positioning errors can often lead to task failure. Therefore, this work contributes by proposing an approach based on Model Predictive Control for footstep planning, a key step for precise humanoid robot locomotion. It examines the theoretical foundations, the constraints of the chosen robot platform, and two techniques for handling obstacles: “Step-over Avoidance” and “Bypass Avoidance”. Additionally, two methods — the “Increasing Horizon Method” and the “Constraint Relaxation Method” — are introduced to find minimum time solutions. The paper also details data collection, performance analysis, and presents the results and conclusions of the developed program.