A model predictive control strategy to regulate movements and interactions

Humans are adept at moving the arm to interact with objects and surfaces. The brain is thought to regulate motion and interactions using two different controllers, one specialized for movements and the other for force regulation. However, it remains unclear whether different control mechanisms are necessary. Here we show that the brain can employ a single high-level control strategy for both movement and interaction control. The Model Predictive Control (MPC) strategy introduced in this paper uses an internal model of the environment to plan the arms muscle activity whilst updating its predictions using periodic feedback. Computer simulations demonstrate MPCs ability to produce human-like movements and after-effects in free and force field environments. It can simultaneously regulate both force and stiffness during interactions, and can accomplish motor tasks demanding transitions between motion and interaction control. Model Predictive Control promises to be an important tool to test ideas of motor control as it can handle nonlinear dynamics with changing environments and goals without having to specify the movement duration.