Interview With Prof. Daniel Ferris
Professor Daniel Ferris is a world reknown expert on the subject of exoskeletons. Dan's impressive credentials are outlined at the bottom of this page. Dan is well published on the topic of exoskeletons as can be seen from a search in Google Scholar. One of the things that inspired Dan in his professional career was his experience as a football player at the University of Central Florida. On the gridiron Dan broke both legs, his hand, nose, and back all at different times and wanted to build a better machine which would not break so easily. We conducted this interview by email one question at a time. It took several weeks due to Dan's busy schedule. We completed the interview in January 2011. Gene: Of the leading exoskeletons (Rex, ReWalk, HAL, eLEGS, etc) which do you believe is technically superior? Why? Dan: Unfortunately, it is not feasible to answer that question at present. It would be like answering “which galaxy has the most living organisms in it?” There is a clear answer, but we don’t have sufficient data to know. The same is true of the exoskeletons. There are lots of press releases, media interviews, and other forms of advertisements in the public domain, but there is very little hard data evaluating their abilities. From the limited data that is available, I would venture a guess, and it really is a guess, that the Honda exoskeletons, HAL, and eLEGs are the farthest along at this point. It is important to point out though that each has different goals and limitations. Thus, each may find a unique niche in the market. Gene: Each exoskeleton manufacturer has used different mechanism for control of the computer which drives the exoskeleton, from a joystick (Rex) to electromyography sensors (HAL) to force and motion sensors (ReWalk and eLegs). Can you speak of the advantages and shortcomings of each? Dan: I am major proponent of embodied control. Simply put, that means controlling an exoskeleton such that the nervous system thinks of the exoskeleton as a part of the user’s body. A joystick could never allow the nervous system to process control commands for movement without requiring the brain to do extra calculations to convert joystick movement to body movement. Compared to a natural movement, the joystick requires extra thought processes. This means that a joystick is not likely to provide the same smoothness and skill that embodied controllers should be able to. On the other hand, a joystick has the benefit that it requires little training to produce a functional movement. Electromyography sensors, such as in the HAL exoskeleton, record the electrical activity across a muscle that enable it to contract. With these sensors, it is possible to activate the exoskeleton in a scaled response to the magnitude of the human muscle activity. The more human muscle recruited, the more exoskeleton assistance. This type of control seems to produce a high level of embodiment. The users implicitly learn to control the exoskeleton without any explicit understanding of how the control is happening. They adapt their own muscle recruitment patterns to take into account the exoskeleton dynamics. The biggest downside to this type of control is that the sensors used for the control are error prone in that they have a poor signal to noise ratio. This makes them unreliable. Also, in many patient populations, there is not a robust muscle activation signal to record for control. Patients with spinal cord injury or stroke likely couldn’t use this type of control very well. The most common type of control strategy employed for exoskeletons in development is using force and motion sensors to decipher the user’s movement intent, such as as with eLegs. Based on the sensor data, the computer controller activates the exoskeleton to assist the movement. The advantages of relying on force and motion sensors for control are that the sensors that are typically very reliable and the software and hardware are relatively easy to develop. The biggest unknown with using force and motion sensors for control is that it is not clear how smooth and accurate the exoskeleton can be working in tandem with a human. We can’t write accurate equations describing how the human nervous system is working under most conditions. Thus, there is a certain amount of reacting that is required using force and motion control. That tends to slow down the system and make the exoskeleton less smooth and efficient. There isn’t any good data yet on the level of embodiment produced by these force and motion controllers. My guess is the force and motion controllers can be good in the best case scenario, but will never reach a level of true embodiment. Gene: At least one exoskeleton eLEGS, is emphasizing the development a more physiological gait. How important is gait in the function of an exoskeleton? If the user is paralyzed, do they really care if the gait is physiological or not? Dan: The word “gait” means different things to different people. In this instance, I believe, gait refers to a kinematic pattern similar to walking by neurologically intact individuals. Kinematics are what you see with your eyes about the motion, without regard to the forces that are causing the motion. Individuals with disabilities actually care quite a bit about how their movements are perceived by others. Increasing mobility with a device that emphasizes abnormal movements and calls a lot of attention to the user would be less desirable to individuals with disabilities than a device that is less noticeable. Wearing any exoskeleton is likely to call some attention to a user, but if it produces what are perceived as normal movements that would be good from the perspective of the user. From a rehabilitation or physiology standpoint, we don’t really know if there is a benefit from having a more normal kinematic pattern. Gene: Most of the leading exoskeletons require that the user has upper body strength and use crutches to support the exoskeleton. This means that most quadriplegics are excluded. The only exoskeleton that I am aware of that does not require the use of crutches is Rex. Since almost half of spinal cord injury results in quadriplegia, this is very significant. Will the requirement for upper body strength and crutches be a major hurdle for the other exoskeletons to overcome in order to be adapted for use by quadriplegics? Is there hope that quadriplegics will also be able to use these kinds of exoskeletons one day? Dan: The number of individuals with mobility impairments from stroke, traumatic brain injury, cerebral palsy, and multiple sclerosis far exceeds the number of individuals with mobility impairments from spinal cord injury. As such, there is a large population of potential exoskeleton users that could rely on crutches when wearing the exoskeletons. However, having acknowledged that, working lower limb exoskeletons that require the use of crutches is just the first step of what is coming. With advances in exoskeletons that require crutches, it is highly likely that future exoskeletons will be able to be used without crutches. Gene: Most exoskeletons seem to have been developed to help paraplegics. Is the technology significantly different for other types of disabilities, such as spasticity related disabilities, cerebral palsy, spastic paralysis, multiple sclerosis, etc as opossed to paralysis? That is, can one exoskeletons work for most people with mobility disorders? If not, which type of exoskeletons do you think would work best with these types of disabilities? Dan: The easiest disability to counteract with an exoskeleton is weakness. However, many neurological disabilities come with other symptoms such as spasticity. Spasticity is harder to deal with because it produces fast, unpredictable motions that would have to be dampened by the exoskeleton. Based on studies that have been done in the past using robotic interfaces to study spasticity, an exoskeleton that can handle strong levels of spasticity is going to be harder to achieve. Given the range and magnitude of symptoms that can be present in patients with the same neurological conditions, it is better to think about the individual fit for an exoskeleton rather than the fit based on the condition itself. There will be a need for different types of exoskeletons, or at least different controllers, for different individuals due to their specific abilities and symptoms. Gene: I have seen predictions that exoskeletons will decrease the risk of hospitalization of wheelchair users due to diseases of the uninary tract, skin (pressure sores) and respiratory system. Others talk about reduced osteoporosis and diseases of the cardiovascular system and the digestive system. What is your take on that? If there are less requirements for hospitalization, can exoskeletons improve the health of wheelchair users, increase lifespan or save money due to less expenditure on health care? Dan: The logic for that prediction, reduction of secondary medical complications with increased exoskeleton use, is sound but we don’t have data to back it up yet. The idea is that reducing sitting time will lead to lead blood flow and tissue breakdown problems. Standing and walking will increase circulation and presumably lead to increased heart rates and respiration compared to sitting in a wheelchair. The loading on the legs should help counteract osteoporosis because bones get stronger when you load them. I think there is a high probability this will occur, but we won’t really know until we have exoskeletons in use enough during the day to make the measurements. Gene: When, or for whom, might the use of an exoskeleton be harmful or not advisable? Dan: For an individual that has some ability to walk on their own, too much use of an exoskeleton could actually lead to atrophy of their own muscles. For other individuals, they would need to the ability to control and drive the exoskeleton safely. Unless the exoskeleton has some balance and perturbation response abilities, then the user needs to be able to safely keep their balance. This ability will be different depending on the exoskeleton.
The advantages of relying on force and motion sensors for control are that the sensors that are typically very reliable and the software and hardware are relatively easy to develop.
Standing and walking will increase circulation and presumably lead to increased heart rates and respiration compared to sitting in a wheelchair. The loading on the legs should help counteract osteoporosis because bones get stronger when you load them. I think there is a high probability this will occur.
Prof. Ferris At TEDx
Gene: What about a person who has been in a wheelchair for many, many years? Could there be a danger that their bones are too brittle, their ligaments too tight or their cardiovascular system too weak to handle the stress of standing again? Or would it be generally possible help most people make this transition?
Dan: Use of the exoskeleton would still be beneficial to them, but it would require graded increases in the amount and intensity of physical activity with the exoskeleton. Milo of ancient Greece started lifting and walking around with a calf on his shoulders every day. When the calf had become a 4-year old Bull, Milo was able to still walk around with the bull on his shoulders. The same sort of progressive training is a critical principle of strength training and rehabilitation. Using the exoskeleton a little more each day would allow the individuals to strengthen their muscles, bones, ligaments, and cardiovascular systems. The rate of increased use would be specific to each individual.
Gene: Let's say that a user, for example an incomplete paraplegic, has some muscle control. For example say they have some function in the hips and upper legs. Lets assume that this person wants to use the exoskeleton to get up and exercise. Shouldn't the exoskeleton be able to do this? I asked one manufacturer and they said that their exoskeleton was not capable of this. Why should this type of "power sharing" function not be optional in an exoskeleton? Is this function technologically difficult to achieve?
Dan: It will totally depend on the control system of the exoskeleton. Some will be able to do it, others won’t. It is doable, but only so far as the exoskeleton control system allows. The short answer is that it depends on what types of sensor inputs are important to the exoskeleton control algorithm. If the exoskeleton control algorithm can’t sense and accommodate biological torques at the joints caused by muscle, then the controller will be perturbed by any active user effort.
Gene: Are there some specific exoskeletons that you would think should allow this kind of "power sharing"? If so, which ones?
Dan: It is hard to say because there is so little data out there on the exoskeleton control systems. However, based on what is available, eLegs, HAL and the Honda exoskeletons should all be able to perform this type of power sharing between the exoskeleton and user.
Gene: Most exoskeletons currently require a backpack. Is that mostly due to the batteries? What else is in the backpack? Are there any major breakthrough in batteries which could lead to a less bulky exoskeleton and perhaps no more need for a backpack? Are the batteries potentially harmful?
Dan: The batteries are usually the part of the exoskeleton that requires the most mass. Battery technology is getting better but it isn’t at a level yet where it has the ideal power density. Some of the computing hardware could go into a backpack, but the small size of computer chips get smaller and other electronics makes it possible to locate the computing hardware elsewhere. There shouldn’t be any harmful aspects from the batteries for most consumer uses. They only become a problem in industrial (think firefighters going into a burning building) or military settings.
Gene: Currently, I am aware of two types of actuators (the motors that drive the exoskeletons), hydraulic and electronic. Are there others? Which are most efficient or better in an exoskeleton.
Dan: Electromagnetic and hydraulic actuators are the most used actuators for current exoskeletons in development. However, there are exoskeletons that use pneumatic actuators, electrorheological actuators, and elastic springs to store and return elastic energy. A recent trend has been to develop hybrid actuators that use combinations of these types of actuators. They each have advantages and disadvantages, so hybrid actuators can overcome some of the biggest disadvantages of individual actuators. There are a host of other actuators in development that may be ready for exoskeleton use in 10 or more years (shape memory alloys, piezoelectric actuators, etc.). Efficient is a very specific measurable parameter for an actuator (mechanical energy produced by the actuator divided by energy consumed by the actuator), and efficiency isn’t always the most important factor in determining the success of an exoskeleton actuator. The type of exoskeleton controller can also influence what type of actuator is best for the exoskeleton. For example, some exoskeleton controllers that provide power sharing benefit from having mechanical compliance inherent to pneumatic actuators. Other exoskeleton controllers depend on very fine position control, so they would benefit more from using hydraulic actuators.
Gene: Given that exoskeletons are just starting to be marketed now and that costs are are currently too high for most wheelchair users to afford, would you expect to see exoskeletons for wheelchair users widely available in the short term (2-5 years?), mid-term (5-10 years) or do you think we should look for them to be made widely available in the longer term (10+ years)?
Dan: Definitely the longer term. Even the ones that are widely reported in the media right now are not achieving the desired performance necessary for widespread market availability
Gene: Can you give us an overview on the kind of projects you are working on which involve exoskeletons?
Dan: The main purpose of my research on robotic exoskeletons is to reveal general principles about how humans respond and interact with mechanical assistance to human locomotion. My lab’s primary goal is not to build commercial products, but to conduct basic science using robotic exoskeletons to perturb and assist human walking and running. By understanding how humans respond metabolically, biomechanically, and neurophysiologically to mechanical forces from robotic exoskeletons, we can suggest ways that better robotic devices can be built that truly benefit human movement.
Dr. Ferris is an Associate Professor in the Department of Movement Science and Department of Biomedical Engineering at the University of Michigan. He is also a member of the Neuroscience Graduate Program and an Adjunct Associate Professor of Physical Medicine and Rehabilitation. He completed a bachelor’s degree in Mathematics at the University of Central Florida, a master’s degree in Exercise Physiology at the University of Miami, and a doctorate degree in Human Biodynamics at the University of California, Berkeley. After post-doctoral fellowships at the UCLA Department of Neurology and the University of Washington Department of Electrical Engineering, he began as a faculty member at the University of Michigan in Ann Arbor in 2001. His research focuses on the biomechanics and neural control of human locomotion, specifically in regards to basic scientific principles and applied methods for gait rehabilitation after neurological injury.
