Magazine article AI Magazine

Control Strategies and Artificial Intelligence in Rehabilitation Robotics

Magazine article AI Magazine

Control Strategies and Artificial Intelligence in Rehabilitation Robotics

Article excerpt

As robotics moved from industrial to service applications, engineers began looking for new tasks that could be automated with robots. Industrial tasks had been a perfect candidate for automation since they are physically exhausting and require high precision. Motor rehabilitation seemed like a similarly appropriate robotics application. In the course of rehabilitation, the patient must exercise by performing limb motions thousands of times, and the therapist must physically support and guide the patient's limb during these motions. Since therapists inevitably get exhausted, a rehabilitation robot could support and guide the limb instead.

Numerous rehabilitation robots have been designed for both the upper (figure 1) and lower limbs (figure 2). The two most famous arm rehabilitation robots are the MIT-MANUS, now sold as the InMotion ARM (Interactive Motion Technologies, USA) and the ARMin, now sold as the ArmeoPower (Hocoma AG, Switzerland). The most famous leg rehabilitation robot is the commercially available Lokomat (Hocoma AG, Switzerland), with another notable example being the Gait Trainer (Reha-Stim, Germany). All of these, and many other robots, were developed in order to support and guide the patient's limbs. However, appropriate hardware is not enough; both therapists and robots need to intelligently adapt their support to ensure proper exercise. Mistakes should be corrected, but the patient should exercise actively and intensely, so the support should not be excessive.

The first rehabilitation robot controllers did not adapt their support to the patient at all. They were very stiff, and essentially guided the patient's limbs along a predefined trajectory with little care for what the patient was doing or wanted to do. Clinical tests found that patients put significantly less effort into robot-aided exercise with such controllers than into therapist-aided exercise, and frequently just let the robot move their passive limbs without actively participating in the motion (Israel et al. 2006, Ziherl et al. 2010). This "slacking" process leads to slower neuromotor recovery (Casadio and Sanguineti 2012). To avoid it, the robot needs to adopt a control strategy that assists the patient only as needed: a cooperative control strategy.

Help Me Help You: Cooperative Assistive Control

Assistive controllers are the dominant control paradigm in rehabilitation robotics, and are used in the majority of commercial systems. They operate on the level of the individual motion, helping the patient complete a motion within a desired time while correcting any major errors (such as large deviations from an optimal trajectory). The main characteristic of modern assistive controllers is that they only help as much as it is necessary for the patient to complete a motion, an approach called patient-cooperative control (Riener et al. 2005). This is similar to the work of therapists in rehabilitation: they manually move the patient's limb to accomplish a desired motion, but let the patient move on his or her own whenever possible.

As summarized by Marchal-Crespo and Reinkensmeyer (2009), many rationales have been given for such assistive controllers. Aside from allowing patients to perform more movements in a shorter amount of time, they interleave active effort by the participant with stretching of the muscles and connective tissue, they provide novel somatosensory stimulation that helps induce brain plasticity, and they may help teach patients to perform demonstrated patterns. Although most of these rationales have not been extensively clinically verified (Marchal-Crespo and Reinkensmeyer 2009), assistive control algorithms remain dominant, particularly impedance-based control.

Impedance-Based Assistance

The cooperative principle of impedance-based controllers is as follows: while a patient is moving along a desired trajectory, the robot does not intervene, but it corrects deviations from this trajectory by applying a force to the patient's limb. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.