Neurological damage often results in motor impairments and reductions in individualsâ functional abilities, particularly when the upper limbs are affected. Many individuals fail to achieve potential recovery of hand function due to insufficient volume of arm-hand training. Robotic rehabilitation has been proposed as an adjunct to physical therapy to meet this shortfall. However, it may be the case that functional recovery is also limited by practice. It has been observed that advances in motor control theory, mostly arising from computational modelling of behavioural data, have not been integrated in rehabilitation practice. The relatively new field of rehabilitation robotics presents a perfect application and testbed for principles of motor theory. Moreover, one specific motor theory, perceptual control theory (Powers, 1973, 2008), has been identified as architecture for designing robotic control architectures (Young, 2018). In this thesis, we commence a research agenda that aims to test the proof-of-principle of PCT to inform robotic motor rehabilitation. The thesis consists of five original research reports. Two systematic reviews were conducted. The first aims to establish whether end-effector devices for arm and hand rehabilitation are efficacious in reducing impairment and improving functional outcomes. Findings suggested that device training may be efficacious for acute, subacute and chronic stroke patients. Thus it was determined that a research agenda which aimed to inform device development through motor theory was justified. A second systematic review evaluated the research literature regarding PCT models of manual tracking performance, in order to determine the extent to which PCT can account for motor performance in the task. Several key limitations in the PCT modelling literature were found. These limitations were investigated in a series of tracking experiments in which PCT models were optimised to, and simulated individual performance. In the first experiment, we developed a test for model individual-specificity. This was applied to PCT models and it was found that optimised PCT models simulated performance at validation (one-week later) with a higher degree of accuracy than a general PCT model. This demonstrates that the PCT model can discriminate between individual control characteristics. In the second experiment, we aimed to establish the effect of delay on model performance as it was not clear whether PCT models could compensate for long feedback delays that are present in the central nervous system. Four PCT model architectures were compared. The standard PCT position control model showed a reduction in model fit to anticipatory tracking behaviour at increasing delays. Conversely, models that controlled a novel perceptual variable (integrated motion representations) showed no such cost to performance at longer delays. Thus, given the appropriate controlled perceptual variable, PCT models can compensate for sensorimotor delays in motor performance. In the final experiment we aimed to investigate the generalisability of the model to a different task conditions. We evaluated whether the most superior model from the previous experiment could make individual-specific predictions (as per the first experiment), when individuals tracked sinusoidal and pseudorandom targets at different speeds with a new apparatus (steering wheel). The model was found to generalise well across task constraints. In sum, the thesis develops a control model that can characterise and simulate individual performance over a range of task constraints. In the process, we addressed several important limitations in the evidence base for PCT. The implications of these advancements for the fields of motor control, and rehabilitation, are discussed in the concluding chapter. The aim of future work will be to implement the novel PCT architecture within the control algorithm for a robotic device for motor rehabilitation.
- Perceptual control theory
- Manual tracking