Identifying time-varying neuromuscular response: experimental evaluation of a RLS-based algorithm

Methods for identifying neuromuscular response commonly assume time-invariant neuromuscular dynamics. However, neuromuscular dynamics are likely to change during realistic control scenarios. In a previous paper we presented a method for identifying time-varying neuromuscular dynamics based on a Recursive Least Squares (RLS) algorithm. To date, this method has only been validated in a Monte Carlo simulation study. This paper presents an experimental validation of the same method. In the experiment, three different disturbance-rejection tasks were performed: a position task with the human instructed to minimize the stick deflection in front of an external force disturbance, a relax task with the instruction to relax the arm, and a time-varying task with the instruction to alternate between position and relax tasks. The position and relax tasks induce different time-invariant neuromuscular dynamics, whereas the time-varying task induces time-varying neuromuscular dynamics. The RLS-based method was used to estimate neuromuscular dynamics in the three tasks. The neuromuscular estimates were reliable both in time-invariant and time-varying tasks. These findings indicate that the RLS-based method can be used to estimate time-varying neuromuscular responses in human-in-the loop experiments.