Optimal control of FES-assisted standing up in paraplegia using genetic algorithms

• Published in: Medical engineering & physics 1999-11, Vol.21 (9), p.609-617
• Main Authors: Davoodi, Rahman, Andrews, Brian J.
• Format: Article
• Language: English
• Subjects: Algorithms; Biological and medical sciences; Biomechanical Phenomena; Calibration; Computer Simulation; Control system synthesis; Controller; Diseases of the nervous system; Electric Stimulation Therapy - methods; Functional electrical stimulation; Fuzzy Logic; Fuzzy sets; Genetic algorithms; Hip Joint - physiopathology; Humans; Joints (anatomy); Knee Joint - physiopathology; Medical sciences; Models, Biological; Optimal control systems; Paraplegia; Paraplegia - physiopathology; Paraplegia - rehabilitation; Posture; Radiotherapy. Instrumental treatment. Physiotherapy. Reeducation. Rehabilitation, orthophony, crenotherapy. Diet therapy and various other treatments (general aspects); Sensitivity and Specificity; Standing up; Three term control systems
• ISSN: 1350-4533; 1873-4030
• DOI: 10.1016/S1350-4533(99)00093-4

A practical system for Functional Electrical Stimulation (FES) assisted standing up in paraplegia should involve only a minimum of manual set up and tuning. An improved tuning method, using a genetic algorithm (GA) is proposed and demonstrated using computer simulation. Specifically, the GA adjusts the parameters of fuzzy logic (FL) and gain-scheduling proportional integral derivative (GS-PID) controllers that electrically stimulate the hip and knee musculature during the sit–stand maneuver. These new GA designed controllers were found to be effective in coordinating volitional and FES control according to formulated criteria. The latter was based on the deviations from a desired trajectory of the knee and hip joints and the magnitude of the voluntary upper body forces. The magnitude of the average arm forces were slightly higher when compared with the open-loop maximal stimulation of the hip and knee musculature; however, the terminal knee velocities were significantly reduced to less than 10°/s. For practical implementation, the number of trials required to optimize the FL and GS-PID controllers can be reduced by a proposed pre-training procedure using a computer model scaled to the individual. The GA designed controllers remain near optimal provided the model–subject mismatch is small.