Inferring Human Subject Motor Control Intent using Inverse MPC

The control intent is a strategy which a human subject follows to accomplish a certain motor control task such as controlling the trunk movement. The problem we address is how control theoretic approaches can capture such a strategy taking into account the physiological constraints and which approach is better. We present such an analysis, so-called intent-inferring. The control intent can be inferred by estimating the cost function that is minimized by a control system. We propose an inverse model predictive control (iMPC) algorithm to infer the control intent. We solve the iMPC problem for an illustrative example to evaluate the algorithm and to compare against inverse linear quadratic regulator (iLOR) approach. The simulated results show that our algorithm can recover the true cost function weights, and outperform the iLOR approach for the illustrative example.