Adaptive and Learning Control Applied to Powered Walkers for Disabled Individuals

Period of Performance: 09/01/2015 - 08/31/2016

$225K

Phase 1 SBIR

Recipient Firm

Barron Assoc., Inc.
CHARLOTTESVILLE, VA 22901
Principal Investigator
Principal Investigator

Abstract

DESCRIPTION (provided by applicant): Barron Associates, Inc. (BAI) has teamed with the University of Virginia's (UVA's) Motion Analysis and Motor Performance (MAMP) laboratory to propose development of an advanced control and computer learning strategy that will intelligently drive a powered walker for people with neuromuscular walking disabilities. The aim of the control strategy is to provide powered assistance that optimally reduces the metabolic cost of walking. The overarching goal of the proposed effort is to reduce the workload of walking, keeping this population walking longer, providing critical exercise, continued muscle development, and improved quality of life. Past approaches have collected gait data in individuals with walking disorders through laboratory trials. Using this information, a prescribed periodic locomotive force was applied to the wheels of the walker in such a manner to mimic walking of individuals without disabilities. However, this "open loop" approach falls short due to the substantial variability in gait characteristics between individuals. This is further confounded by the substantial variability in gait in each individual, due to growth, muscle development, fatigue level, walking surface, uphill/downhill, obstructions, etc. BAI proposes to develop a walker with powered wheels driven by an onboard microprocessor that houses an advanced, closed-loop adaptive control strategy to rapidly adjust the input force cycle to account for this variability in real-time and address current variations in user inputs. Accuracy of the adaptive controller is improved with more accurate system models. This is achieved by employing an autonomous composite learning algorithm that models the user and walker as one system and learns typical model characteristics common to the individual. Over time, the microprocessor will save and continually update these models to further enhance the robust performance of the adaptive controller. Such user models represent the typical gait characteristics of an individual, such as when s/he is well rested, tired, traversing an inclined path, etc. In Phase I, the main work tasks are: (1) UVA's MAMP laboratory will first conduct a series of human participant trials to collect gait data of both normally developed and walking disabled persons using in-house unpowered walkers with instrumented hand grips and a motion capture system;(2) This data will then be used to build coupled user/walker simulation models;(3) Prototype control strategies will be designed using these simulation models;(4) The developed control strategies will be ported to a powered laboratory walker and trials performed in individuals with walking disabilities;and (5) The collected data will be analyzed and the efficacy of the control strategie assessed. If needed, further control strategy refinements will be completed in Phase II. The primary focus of Phase II will be to further enhance the control system design, construct a prototype walker with microprocessor and drive motors for the wheels, and conduct expanded human participant trials and evaluations.