Tradeoffs in optimal control capture patterns of human sensorimotor control and adaptation

Abstract

Modern control theory highlights strategies that consider a range of factors, such as errors caused by environmental disturbances or inaccurate estimates of body or environmental dynamics. Here we reveal similar diversity in how humans naturally adapt and control their arm movements. We divided participants into groups based on how well they adapted to interaction loads during a single session of reaching movements. This classification revealed differences in how participants controlled their movements and responded to mechanical perturbations. Interestingly, variation in behaviour across good and partial adapters resembled simulations from stochastic and robust optimal feedback control, respectively, where the latter minimizes the effect of disturbances, including those introduced by inaccurate internal models of movement dynamics. In a second experiment, we varied the interaction loads over short time periods making it difficult to adapt. Under these conditions, participants who otherwise adapted well altered their behaviour and more closely resembled those using a robust control strategy. Taken together, the results suggest the diversity of how humans control and adapt their arm movements may reflect the accuracy of (or confidence in) their internal models. Our findings may open novel perspectives for interpreting motor behaviour in uncertain environments, or when neurologic dysfunction compromises motor adaptation.