Adaptive Aiding for Artificial Limb Control

Studies that were conducted by the UCLA Bidtechnology Laboratory developing techniques for enhancing patient control for arm prostheses have produced a variety of upper body muscle transducer control systems, which were reported in the Bulletin of Prosthetics Research (1) and laboratory reports (2,3) . In pursuing this work it was revealed that there is a basic limitation on the ability of a patient to supply the control information rate required to obtain adequate levels of performance (4) . This pgper describes a technique to solve the information problem by adding ah aiding subsystem to the control loop. In such configuration the decision load of the patient is shared between the patient and the aiding subsystem. In an attempt to examine the feasibility of active aided control, an aiding subsystem was developed and demonstrated at the Biotechnology Laboratory. The system utilizes upper body transducer control and a 4 deg.-of-freedom bench model prosthesis. It incorporates an Autonomous Control Subsystem (ACS) able to supplement the patient's control function while operating in parallel with him. The behavior of the ACS is established through a learning process which involves observing. the patient's control function in relation to the environment and task requirements. The computer-based system establishes a decision-making policy which utilizes conditional probability. Initially, the artificial arm is controlled by the patient and the computer system is a passive observer. As the operation continues, the computer system is gradually transformed to an active controller, thereby reducing the patient's function to that of an action initiator and inhibitor. A learning model based on maximum likelihood principles constitutes the mathematical basis for the system.