Learning to Control Complex Rehabilitation Robot Using High-Dimensional Interfaces

Abstract

Learning to perform everyday tasks, using a complex robot, presents a nested problem. It is nested, because, on the surface, there is a problem of robot control---but within it, there lies a deeper, more challenging problem that demands the control nuances necessary to perform complicated functional tasks. For individuals with limited mobility, such as those with cervical spinal cord injuries, the addition of physical burden is added to this motor learning burden. An explicit training regime can be designed to accelerate and aid the learning process, with the long-term aim to help individuals (injured or uninjured) acquire the skill of complex robot control. However, such training regimes are not well-established nor are the methods of evaluation. In this paper, we gain a baseline understanding of how humans learn to control a 7 degree-of-freedom assistive robotic arm, using a novel high-dimensional interface, in the absence of explicit training. We examine how participants transition between distinct workspace zones to extract their learning possibilities. We gain additional granularity in individual learning, based on how participants spend their time in the workspace, with the robot, and how the time spent is distributed across trials. These analyses highlight the high diversity of learning. Lastly, we provide benefits and opportunities for targeted training regimes that are explicit and heavily favor individualized support.

Embodied Communication: How Robots and People Communicate Through Physical Interaction

“…Motion can also be used to purposefully communicate intent, e.g., through gestures (31), motion cues (32), facial movements (33), shoulder shrugs (170), or gaze (34,35). Substantial progress has been made in sign language recognition that could be used in human-robot collaboration (36).…”

“…If we are mapping joystick commands to the action space of a powered wheelchair, there is an intuitive way to specify the mapping of the continuous 2D control space onto the analogous 2D action space of the robot, making the language easy to learn. However, if we are using shoulder shrugs to control movements of a robot arm (170) or sips and puffs to maneuver a wheelchair (58), the mapping is no longer as intuitive, either to specify or to learn, and may not even be within the motor control capacity of a given individual.…”

A Non-Linear Body Machine Interface for Controlling Assistive Robotic Arms

“…Assistive devices such as powered wheelchairs and robotic manipulators can return independence to people with motor impairments. However, controlling a mechanism with multiple DoFs is a specific challenge [11], [30]- [32]. Commercially available interfaces can ease the users' burden of controlling the device by making use of a discrete control system.…”

“…First of all, we wanted all participants to start from a common initial condition to ensure a fair comparison of their learning curves during the operation with the interface. Our group has used this design in previous studies as well [18], [46], [47]. Secondly, a subject-specific mapping would have been inconvenient to implement in real time, as the process of AE validation requires to train several AE models, which is time consuming.…”

“…BoMIs have been tested in situations involving the control of a computer cursor [8], a powered wheelchair [9], a robotic manipulator [10], [11], and quadcopters [6], [12]. They most often rely on a linear dimensionality reduction (DR) technique such as Principal Component Analysis (PCA) [13].…”