Obstacle Avoidance in Intelligent Assisted Walking Devices for Improving Mobility Among Seniors with Cognitive and Visual Impairments

Abstract

Current research in walkers and rollators with integrated intelligent computing and robotic components shows promise in treatment, management and rehabilitation of a variety of ailments and disorders such as stroke, Alzheimer disease and multiple sclerosis. In this thesis a novel intelligent walker is designed, constructed and tested for the purpose of examining whether we can increase mobility among individuals with vision and cognitive impairments hindering their ability to move collision free about their environments, by detecting obstacles and using brakes to guide the user around them. This walker consists of a support frame, front castor wheels and rear particle brakes. Obstacle detection and localization are sensed by an onboard 3D depth camera and RGB camera (The Microsoft Kinect) and encoders in the rear wheels. This data is processed by an onboard laptop, producing a 2-dimensional map of the environment. This map is inputted into the control algorithms to make braking decisions for obstacle avoidance. Two control algorithms are presented. The first is an open loop proportional gain control which determines necessary braking torque directly from the distance to the nearest obstacle. The second is a closed loop control which uses the systems dynamics and velocity data from the wheel encoders to estimate the forces being applied by the user and calculates the braking torque necessary to avoid obstacles. The walker moment of inertia and the viscous damping parameters of the system are estimated experimentally. The effect of varying three parameters in the closed loop algorithm and one parameter in the open loop algorithm are examined in a corner turning test. Observations support predictions made by the derived system dynamics. Lastly, the efficiency of the system at real world obstacle avoidance is tested in a controlled indoor obstacle course using goggles to impair the vision of otherwise able bodied test subjects. The open loop control algorithm was found to reduce the occurrence of collisions by 44% as compared to trials with no braking. The closed loop control algorithm was found to greatly reduce collisions with the front of the walker, however shows a tendency for over steering the user, producing a higher number of collisions with the walker's side. Possible causes and solutions to this problem are discussed. This thesis demonstrates promise in the approach of using braking to help walker users avoid collisions with their environments. Discussion is offered about necessary next steps towards testing with regular users of assisted walking devices, and eventually real world use.