Health Care Robotics – PAM
The goal of this project is to create a patient assistant mobile (PAM) robot to monitor patient behavior and provide a cost-effective physical presence to continuously observe a patient and the environment to prevent falls.
The project will advance state-of-the-art prevention and understanding of falls through continuous situational monitoring and real-time mobility assessment using fall protection and prevention algorithms and modeling. Major scientific challenges are development of (1) an intelligent, mobile robotic aide system and predictive models to provide assistance prior to a fall (i.e. bedside transition, sit-to-stand, toileting), (2) a sensor driven system model combining the roles of many known risk factors associated with falls and an immediate response to patient mobility needs through a low-cost hybrid robot, (3) design of and control algorithms for optimizing the delivery and maneuverability of mobility aids during transitions from the bedside or bathroom, and (4) system development to measure fall risk and analyze, communicate and present data to individuals and care providers to prevent falls and reduce the associated injuries from falls.
When resting, PAM will stay in its docking station (pre-set location). When human movement is detected, PAM will determine the level of fall risk and engage as needed. The figure represents how PAM will acquire a mobility aid and bring it to an individual to leave the bedside and ambulate to the bathroom.
A novel actuated gait simulator has been developed that fits inside a typical shoe and has sufficient load capability and range of motion to mimic adult gait patterns. The control system includes passive ankle stiffness which is responsible for a more biomechanical behavior. The ankle demonstrates passive ankle stiffness when irregular terrain is encountered. Beyond the present application of the device as a testing platform for haptic smart orthotics, it can also be used as a test platform for studying slip events between different types of shoes and varying terrain. I order to have more realistic and biomechanical responses, a centralized control system is being developed which will perform based on a biomechanical model of human gait
The goal of this project is to model stiffness of a multi-segment foot during normal gait. We are trying to redesign the foot for the testbed so that more realistic forces are obtained. The idea is that we add joints to the foot that have some amount of stiffness that allow the foot move like a real foot with similar forces produced. Opensim and Matlab has been used to find the best place to put the joint, how much stiffness is required, and how many joints should be used.
Determining direction of movement through low-cost EEG sensors and machine learning
Detecting motor function in EEG is essential for creating brain machine interfaces for those with limited mobility. Machine learning is able to detect basic motor functions from raw EEG. Further, event related desynchronization can increase detectability of motor related actions. However, brain machine interfaces are limited in function and portability due to EEG motor detection rates, complex EEG-based algorithms and computationally heavy algorithms. By reducing these barriers, EEG-based control can be portable, simplistic, accessible and implemented into the control of upper extremity exoskeleton.
This project will investigate the detection and prediction of unilateral and bilateral motor function of the upper extremity in physical movement and imagery tasks by using light-weight algorithms. We propose that Independent Component Analysis and machine learning techniques can be used to automatically detect basic motor functions, movement direction and muscle activation of the upper extremity. Motor events can be assessed automatically and marked for automatic detection. Event related desynchronization can be observed using spectral band power. Finally, this project will aid in the control of an upper-extremity exoskeleton.