CARRT-Motion Capture Data for Robotic Human Upper Body Model.

In recent years, researchers have focused on analyzing humans' daily living activities to study various performance metrics that humans subconsciously optimize while performing a particular task. In order to recreate these motions in robotic structures based on the human model, researchers developed a framework for robot motion planning which is able to use various optimization methods to replicate similar motions demonstrated by humans. As part of this process, it will be necessary to record the motions data of the human body and the objects involved in order to provide all the essential information for motion planning. This paper aims to provide a dataset of human motion performing activities of daily living that consists of detailed and accurate human whole-body motion data collected using a Vicon motion capture system. The data have been utilized to generate a subject-specific full-body model within OpenSim. Additionally, it facilitated the computation of joint angles within the OpenSim framework, which can subsequently be applied to the subject-specific robotic model developed MATLAB framework. The dataset comprises nine daily living activities and eight Range of Motion activities performed by ten healthy participants and with two repetitions of each variation of one action, resulting in 340 demonstrations of all the actions. A whole-body human motion database is made available to the public at the Center for Assistive, Rehabilitation, and Robotics Technologies (CARRT)-Motion Capture Data for Robotic Human Upper Body Model, which consists of raw motion data in .c3d format, motion data in .trc format for the OpenSim model, as well as post-processed motion data for the MATLAB-based model.

Biomimetic Approaches for Human Arm Motion Generation: Literature Review and Future Directions.

In recent years, numerous studies have been conducted to analyze how humans subconsciously optimize various performance criteria while performing a particular task, which has led to the development of robots that are capable of performing tasks with a similar level of efficiency as humans. The complexity of the human body has led researchers to create a framework for robot motion planning to recreate those motions in robotic systems using various redundancy resolution methods. This study conducts a thorough analysis of the relevant literature to provide a detailed exploration of the different redundancy resolution methodologies used in motion generation for mimicking human motion. The studies are investigated and categorized according to the study methodology and various redundancy resolution methods. An examination of the literature revealed a strong trend toward formulating intrinsic strategies that govern human movement through machine learning and artificial intelligence. Subsequently, the paper critically evaluates the existing approaches and highlights their limitations. It also identifies the potential research areas that hold promise for future investigations.

Enhanced Control to Improve Navigation and Manipulation of Power Wheelchairs.

Our team developed a mobile wheelchair control kit designed to allow power wheelchair users the ability to maneuver their wheelchair without the need to manipulate a joystick with their hands. A smartphone and its internal accelerometer sensor was used to detect the vector of gravity, and thus detecting the pitch and roll. The wheelchair control system was tested with the phone attached in three mounting positions: hand held, hat and arm band and compared to the manipulation using the wheelchair joystick. To determine the viability of the commercialization of this kit as well as which features to further develop, a customer discovery was completed. Over a hundred interviews of power wheelchairs users, therapists, care takers, manufacturers, dealers, and assistive technology professionals were conducted at clinics, tradeshows, disabilities support groups, and rehabilitation organizations. After discovering the needs of the customers, collision avoidance was implemented into the control kit and back up cameras were added into the smart phone app to allow for the camera view to be seen without additional screens attached to the wheelchair. Future work will test these new design features and will concentrate on removing excess weight from the control kit. Improving the ease of installation of the kit to any power wheelchair will also be a focus.

Robotic assistance for performing vocational rehabilitation activities using BaxBot.

Activities of Daily Living (ADL's) refer to tasks that people do on a daily basis, such as self-feeding, cleaning the house, or bathing. These activities often require a degree of functional mobility that may be outside the ability of a person suffering from cognitive or physical impairment. This work describes methods of performing ADL's with a mobile robotic system. We examined the needs of potential users and caregivers through surveys to determine the most needed applications for robotic assistance. Using this information, we extended the functionality of our BaxBot mobile robotic system to provide meaningful, autonomous assistance in performing three specific ADL's with minimal user interaction.

Semi-autonomous mobility assistance for power wheelchair users navigating crowded environments.

Power wheelchair users suffering from cognitive or physical impairment often face difficulties in maneuvering their wheelchairs through crowded environments. Currently, users need to be continuously aware of all traffic around them to actively avoid all collisions. This is an especially difficult task since many wheelchair users are unable to accurately view or perceive their surroundings. Additionally, imprecise joystick control, slowed reaction time, or imperfect interpretation of the environment can lead to unintended collisions with objects in the environment. This work looks to augment user's input with data gathered from an ultrasonic sensor ring to prevent accidental collisions. Using data gathered from the sensors, we detect objects within a certain radius of the chair. This sensor information is combined with the user input from a joystick to generate a potential field description for the intended motion of the wheelchair. An optimal motion vector is calculated which works to avoid collision with obstacles. Ultimately, this control method reduces the cognitive load on the user and enables them to navigate complex environments by providing simple and/or imprecise input to the system.

Human motion intention based scaled teleoperation for orientation assistance in preshaping for grasping.

In this paper, we present an algorithm that provides human motion intention based assistance to users teleoperating a remote gripper for preshaping over an object in order to grasp it. Human motion data from the remote arm is used to train a Hidden Markov Model (HMM) offline. During the execution of a grasping task, the motion data is processed in real time through the HMM to determine the intended preshape configuration of the user. Based on the intention, the motion of the remote arm is scaled up in those orientation directions that lead to the desired configuration, thus providing the necessary assistance to the user to preshape for grasping. Tests on healthy human subjects validated the hypothesis that the users are able to preshape quicker and with much ease. Average time savings of 36% were obtained.

Probability density based gradient projection method for inverse kinematics of a robotic human body model.

This paper presents the probability density based gradient projection (GP) of the null space of the Jacobian for a 25 degree of freedom bilateral robotic human body model (RHBM). This method was used to predict the inverse kinematics of the RHBM and maximize the similarity between predicted inverse kinematic poses and recorded data of 10 subjects performing activities of daily living. The density function was created for discrete increments of the workspace. The number of increments in each direction (x, y, and z) was varied from 1 to 20. Performance of the method was evaluated by finding the root mean squared (RMS) of the difference between the predicted joint angles relative to the joint angles recorded from motion capture. The amount of data included in the creation of the probability density function was varied from 1 to 10 subjects, creating sets of for subjects included and excluded from the density function. The performance of the GP method for subjects included and excluded from the density function was evaluated to test the robustness of the method. Accuracy of the GP method varied with amount of incremental division of the workspace, increasing the number of increments decreased the RMS error of the method, with the error of average RMS error of included subjects ranging from 7.7° to 3.7°. However increasing the number of increments also decreased the robustness of the method.