Fast Model Predictive Control for Torque-Controlled Humanoid Robots

In this article, we present a new approach to robot SRBD (Suspension-Rotation-Bearing Dynamics) modeling that simplifies the dynamics of bipedal and humanoid robots while maintaining their essential dynamic effects. Our method combines the weight of the object being manipulated with the robot’s SRBD model, allowing the robot to compensate for added mass and inertia within the SRBD formulation without sacrificing linearity.
To create a more agile and efficient robot, we also introduce two sets of torque and speed settings. These settings enable the robot to maintain its agility while increasing its torque capacity, making it suitable for various tasks.
Our approach uses timing belts instead of linkage systems, which are commonly used in robotics due to their simplicity but lack of mechanical advantage without introducing nonlinearity that complicates control. Timing belts, on the other hand, provide a more robust and compact solution for transmitting power over extended distances while maintaining linearity in the dynamics formulation.
By combining these innovations, our modular and compact robot SRBD model enables bipedal and humanoid robots to perform a wide range of tasks with improved efficiency and agility without compromising their ability to handle heavy loads or changing environments.