Model-Based Control of Individual Finger Movements for Prosthetic Hand Function

Abstract

Prosthetic devices for hand difference have advanced considerably in recent years, to the point where the mechanical dexterity of a state-of-the-art prosthetic hand approaches that of the natural hand. Control options for users, however, have not kept pace, meaning that the new devices are not used to their full potential. Promising developments in control technology reported in the literature have met with limited commercial and clinical success. We have previously described a biomechanical model of the hand that could be used for prosthesis control. The goal of this study was to evaluate the feasibility of this approach in terms of kinematic fidelity of model-predicted finger movement and the computational performance of the model. We show the performance of the model in replicating recorded hand and finger kinematics and find average correlations of 0.89 between modelled and recorded motions; we show that the computational performance of the simulations is fast enough to achieve real-time control with a robotic hand in the loop; and we describe the use of the model for controlling object gripping. Despite some limitations in accessing sufficient driving signals, the model performance shows promise as a controller for prosthetic hands when driven with recorded EMG signals. User-in-the-loop testing with amputees is necessary in future work to evaluate the suitability of available driving signals, and to examine translation of offline results to online performance.

Fig. 1.

OpenSim visualisation of the hand model, showing muscle lines of action and included joints (CMC, MCP, IP at the thumb, MCP, PIP and DIP for the fingers).

Fig. 2.

EMG electrodes were placed over four key muscle areas (top image) allowing independent control of the five postures (including rest). These were: the lateral part of EDC, the medial part of EDC, the FDS (just distal to the superficial wrist flexors), and the EPB. The middle row of images shows the locations of markers for the kinematic analysis, and the bottom row shows the four target postures presented.

Fig. 3.

Schematic of simulated object gripping. The weight of the cup is altered to simulate filling, and the resultant shear force, F_shear, on the finger is used to estimate the minimum contact force, F_min, necessary to prevent slip. Muscle excitations, u, are input to the model and the resultant finger displacement, x, is output, from where cup stiffness is used to estimate fingertip contact force, F_norm. A PID controller modulates the muscle excitation to ensure the contact force is kept above the minimum.

Fig. 4.

Example of the muscle excitation signals used as model inputs, together with the raw EMG signals from which they are derived. The segment shown includes the ‘thumbs up’, ‘hand open’, ‘pointing with index finger’ and ‘L-shape’ postures.

Fig. 5.

The normalised angles for both the measured hand kinematics and the model-predicted joint postures. These are for the same segment of data as shown in Fig. 4, normalized using the range estimated from the repeated open-close trial.

Fig. 6.

The model also simulates proprioceptive feedback. This figure shows the muscle spindle primary and secondary afferent firing rates for the same segment of data, alongside the normalised fibre length.

Fig. 7.

The Prensilia IH2 Azzurra robotic hand and the participant’s hand, shown in the various postures encountered during the trial. A video of the control achieved using this hand is available in the Supplementary Material.

Fig. 8.

Panel A shows weight of the cup as liquid is added. Panel B shows activation in the FDPI muscle to ensure that the fingertip force just exceeds the minimum necessary to prevent slipping. Panel C shows the resulting fingertip force, Panel D the force in the finger flexor and Panel E the simulated GTO output.