Testing Rx for medical mechatronics
To develop its "human interactive" robotic devices, Kinea Design relies on an eclectic array of tests that include conventional electronic instruments, homegrown solutions, and "human-in-the-loop" techniques.
By Lawrence D. Maloney, Contributing Editor -- Test & Measurement World, 7/1/2010 12:02:00 AM
|
Evanston, IL—When your task is to design innovative devices that will help a person regain movement after a stroke or will help prevent injury from backbreaking labor, there are no application notes to steer you through the test process. The multidisciplinary team of R&D engineers at Kinea Design address such design challenges as part of their daily workload. The engineers, who specialize in what they call "human interactive mechatronics," must be as creative in designing test setups as they are in blending the sensors, actuators, software, and control systems that form the heart of their motion-oriented inventions.
To ensure the R&D team has the necessary skills, Kinea hires versatile engineers who are as comfortable with electronics and control systems as they are with mechanical design and computing tools like Matlab, Simulink, and LabWindows/CVI.
"We look for engineers who can integrate technologies and deal with tough systems problems," said Ed Colgate, a PhD mechanical engineer who founded the company in 2003, along with Michael Peshkin, the president of the company as well as a PhD mechanical engineer, and David Brown, a PhD and a physical therapist. All three are professors at Northwestern University, just a few blocks from Kinea's design lab.
The skills Colgate described have spawned the creation of new test and measurement devices, both for use by patients and for development of inventions ranging from advanced prosthetics and rehabilitation equipment to a "HookAssist" that removes much of the burden of handling beef carcasses in meat-processing plants (Learn about some of Kinea Design's mechatronics projects outside the medical field in "From meat to fish.").
Where robots meet people
|
"In our assist systems, the robot provides assistance, but it is always the person, not the robot, who initiates and controls the movement," explained Peshkin, who also co-founded an earlier company called Cobotics that specialized in hoists, arms, and other assistive devices for assembly-line workers.
Peshkin called Kinea's "KineAssist" system a prime example of this robot-human interface. Targeted for rehabilitation of stroke victims, the device partially supports the upper-body of patients and protects against falls, freeing the physical therapist to focus on walking and balance training.
KineAssist's chief components include pelvic and torso harnesses, trunk- and pelvic-support mechanisms, and a wheeled base that allows a patient to turn and to move laterally, forward, and backward (Figure 1). The patient's intent for motion is detected by potentiometers and force sensors integrated into the pelvic-support structure. Control algorithms move the motorized base in response to the patient's own forces and motions.
 Figure 1. In the KineAssist system, control algorithms move the motorized base in response to the patient's own forces and motions. Courtesy of Kinea Design. |
In addition, the pelvic-support structure provides a vertical force of up to 150 lbs to "unweight" the patient. It allows the walking patient to bend forward, but includes a Safety Zone feature that provides compliant constraint to protect against falls.
"The biggest challenge was the open-ended nature of the project, as we worked to define the requirements of the design," said Julio Santos-Munné, Kinea's director of operations.
Kinea engineers spent many weeks observing therapists at work in clinical settings, including the prestigious Rehabilitation Institute of Chicago (RIC). During those visits, Santos-Munné and controls engineer Alex Makhlin worked closely with physical therapist Ela Lewis, also a member of the development team.
"It was clear that the therapists did not want robotic technology that would interfere with their hands-on work," said David Brown. "Instead, they wanted an intuitive robot that would enhance their skills, ease the patient's fears of falling, and allow for more challenging therapy work."
|
To test the device, the team made extensive use of finite element analysis (Figure 2) to ensure the structural integrity of major components, such as the pelvic mechanism and mobile frame, which were designed with SolidWorks' 3-D modeling software.
Makhlin also used The MathWork's Matlab software for data analysis and dynamic simulation of certain subsystems, especially the kinematics of the wheels and mobile outrigger. Matlab analyzed signals that came from various sources within the system's embedded computer, including optical encoders integrated with the motors, potentiometers at structural joints, and customized load cells placed throughout the system's five axes to sense user intent.
"The Matlab analysis helped us calibrate the sensors and examine the system's dynamic response," said Makhlin, a mechanical engineer whose master's degree is in control engineering. Other software on the embedded processor allowed him to run frequency response analysis and Bode plots, also displayed in Matlab. In addition, Makhlin used Matlab's System ID tool to construct models of the pelvic and torso subsystems, which helped determine stability margins in the design.
Kinea engineers also stress the importance of "human-in-the-loop" testing, where patients or the engineers themselves evaluate prototypes of the device. "On the KineAssist, I would strap myself into the device to see how well it performed, then maneuver it to my desk to make changes in control algorithms in real time," recalled Makhlin.
Among other examples of "human-in-the-loop" feedback, Brown noted that researchers did EMG (electromyogram) tests to determine a patient's muscle activity both in traditional therapy and while using the KineAssist. And to help determine the therapy activity most needed in an assist device, a special lab equipped with eight video cameras was set up at the RIC. There, patients performed functional tasks both with and without KineAssist on a 5-m walkway, and EVaRT software from Motion Analysis was used to reconstruct body-segment motions for analysis. Result: Walking and balance activities were determined to be the greatest needs for the KineAssist technology.
Value of virtual instruments
The KineAssist project also led Makhlin to develop a virtual instrument called EKG that has served the Kinea team on several mechatronics projects. "Just as a medical EKG allows the doctor to study the electrical activity of the heart, I needed a tool that would allow me to probe the whole range of signals coming into the control system from a complex array of sensors and actuators," Makhlin explained.
The software-based tool comprises two segments. One resides on the embedded system and is written in C++ and runs on QNX. The other runs on an external laptop or desktop and is written in Java. Engineers can run EKG without modification on Windows, Linux, and Unix. One researcher at RIC even uses it on a Mac.
EKG, which has a graphical user interface that shows real-time signal plots, communicates with the device's controls software via an Ethernet connection. From a list on the display, the user chooses a signal to visualize, such as that of a force sensor, and EKG plots the corresponding signal (Figure 3).
 Figure 3. Using its homegrown "EKG" virtual instrument, Kinea engineers can probe the whole range of signals coming into a mechatronics control system from an array of sensors and actuators. Courtesy of Kinea Design. |
"There is no limitation on the number of channels you can graph, and you can gather as much data as you like," said Makhlin. "For example, I might want to look at a signal from a potentiometer as well as a command signal to an actuator to analyze the velocity and position that I am asking an actuator to maintain."
Makhlin added that EKG is a great tool for looking at the dynamically changing variables within an embedded control system. If the variables become too numerous, he collects the data and exports it to Matlab for analysis. What's more, the tool is easy to learn and has been used extensively by physical therapists to collect data from the KineAssist while working with patients.
"EKG is a fundamental tool that I can't imagine not having," said Makhlin. "I use it all the time for development and debugging control systems."
Helping to revolutionize prosthetics
While developing its own mechatronics projects, Kinea also has been an active participant in what many experts consider to be the most sophisticated prosthetics design project ever attempted. Sponsored by DARPA (Defense Advanced Research Projects Agency) and involving more than 30 partners from industry, academia, and government, the $60 million Revolutionizing Prosthetics 2009 project sought to design a bionic limb that closely resembled the look, feel, and movement of a human arm and hand.
|
Kinea has worked on several phases of the project, including an earlier Revolutionizing Prosthetics 2007 project, and has designed innovative devices that include:
• miniature fingertip sensors and haptic devices that deliver a variety of touch sensations to an amputee (Figure 4a),
• a sophisticated "cobot" assembly featuring CVTs (continuously variable transmissions) to actuate motion in the wrist and fingers of a prosthetic hand (Figure 4b), and
• a motorized finger and palm assembly featuring a differential linkage and cycloidal drive system.
For these designs, the Kinea team invented new test equipment and devised a litany of homegrown test solutions. These range from mechanical fixtures with load cells for measuring force and flexure on prosthetic fingers to elaborate computer-controlled testbeds that demonstrate the viability of complex electromechanical systems. Time and again, the team relied on feedback from amputees who used the devices—the all important "human-in-the-loop" dimension.
Take, for instance, the work Kinea did on haptic interfaces in the early phase of the DARPA project. These systems basically consist of a fingertip sensor on the prosthetic hand, a haptic tactor placed on the amputee's chest, and an embedded control system to translate signals from the fingertip sensor to recreate the sensation of touch at the haptic tactor.
For this technology to make the most impact, a person must first undergo targeted reinnervation surgery, pioneered by Dr. Todd Kuiken of RIC. In this procedure, a surgeon reroutes nerves from the residual limb of the amputee to the pectoral muscles both to control a hand prosthesis and to receive tactile sensations from it.
The most advanced fingertip sensors that Kinea designed for DARPA provided several different sensations to the amputee: contact, pressure, vibration, temperature, shear force, and four discrete points of contact. The engineers performed a whole battery of tests to validate each of these sensations and used LabWindows/CVI to create the GUI for displaying the results, such as force or heat flux.
"Our test setups for the haptic interface also involved a lot of human-in-the-loop work," recalled Santos-Munné, "and it was amazing to see the patient's reaction to touching different surfaces, like ribbon cable, Teflon, or sand paper."
Answers from the "Greenbox"
To test the tactor system, Makhlin developed a special piece of test hardware he calls the "Greenbox," which has since been used on many Kinea projects (Figure 5). Running on the QNX real-time operating system on software developed in C++, the instrument performs a variety of testing and verification tasks, such as calibrating load cells, checking the closed-loop response of custom actuators, and serving as a CAN bus master during hardware development. It can also interface with a variety of sensors and has a long list of interfaces, including:
|
• 16 analog inputs with 12-bit resolution,
• 16 digital I/O channels (source sink approximately 20 mA),
• eight analog output channels with 16-bit resolution,
• six quadrature encoder channels,
• six differential amplifiers (with a gain of up to 1000X),
• two linear DC motor amplifiers (peak current of 2 A), and
• four ports: RS-232, RS-422/485, CAN bus, and Ethernet.
"I developed the Greenbox originally to serve as a controller/test system for the tactor," said Makhlin, "but I had a hunch that it would be useful elsewhere, because we use a lot of the same hardware in many of our mechatronics systems, such as an embedded CPU, digital/analog I/O, strain-gage amplifiers, signal conditioning for potentiometers, and CAN boards."
The device's wide array of I/O lets it port to a variety of electronic modules, and its onboard x86 computer allows it to run the EKG tool. "If I am working on a microcontroller project, I can interface the microcontroller board to the Greenbox and then stream the data through the EKG," noted Makhlin.
The Greenbox comes into play often in tactor testing. For example, one test measures the closed-loop frequency response of the tactor while its sensored tip presses against a silicone pad. The Greenbox sends sinusoidal commands to the tactor to measure frequency response, and it sends two types of pulse commands for measuring tapping response. Later, data from the test can be exported to Matlab for analysis.
In other instances, Makhlin uses the Greenbox in combination with a controller and digital force gauge to verify a number of functions, including proper calibration of the tactor's load cells, control of the tactor itself, and proper operation of the position sensors integrated with the tactor's miniature DC brushless motors.
Testbed for credibility
Kinea's creativity in testing also played a big role in expanding its involvement with the Revolutionizing Prosthetics 2009 project, particularly in its proposal to develop the cobot assembly. Located in the forearm area and attached to a prosthetic hand, the 2-lb cobot contains a customized 40-W brushless motor, a single drive shaft, and 15 CVTs that activate 15 high-strength fibers. In effect, the fibers serve as tendons for moving the prosthetic hand and wrist.
"Cobots have basically one source of power, but multiple transmissions that can tap that power for multiple degrees of freedom for your robot," explained Kinea's Eric Faulring, the project's lead mechanical engineer. "So, you save on weight and space."
|
Sounds impressive, but would the concept really work? For a crucial meeting with DARPA and the prime systems integrator, the Johns Hopkins University Applied Physics Lab (APL), Faulring developed in six weeks a three-degrees-of-freedom testbed to demonstrate the concept (Figure 6).
The testbed consisted of the main 40-W motor, the drive shaft, and three spherical CVTs needed to move a single prosthetic finger. Each transmission had a miniature motor to adjust the transmission ratio, a potentiometer, a gear reducer, a tiny spool for pulling a finger tendon, and an encoder to measure that pulling action. The output in this testbed was the finger pushing against a force sensor.
To control the demonstration, Faulring designed a "black box" electronics testbed that included a computer running on the QNX operating system with control code written in C, data-acquisition boards, and brushless DC motor amplifiers for the main drive and for the smaller motors that adjust the transmission ratios. Finally, the testbed had two Kinea-designed boards, one for conditioning and filtering signals from the potentiometer, and the other for conditioning and amplifying strain-gage channels.
During the demonstration, Faulring used the EKG to visualize force signals and position signals from the testbed's control system, displaying them on a laptop's GUI. "No one in the program was familiar with a spherical CVT, and people didn't believe it would work," recalled Santos-Munné. "But this successful demo gave project leaders confidence in our team and its concepts."
Engineering and testing refinements
Not only did Kinea get the go-ahead to complete the full cobot design, but the company moved on to tackle additional design concepts for DARPA. For example, the final Revolutionizing Prosthetics 2009 design included not only Kinea's fingertip sensors but also its modular one-motor finger, actuated by a cycloidal gear box to curl in a natural motion around an object. In addition, Kinea's palm module, designed and tested to withstand a fall, serves as the principal electromechanical interface and enclosure for all the electronic and mechanical components connecting the prosthetic hand and wrist.
Integrating its design with those of other companies under APL's direction also enhanced Kinea's expertise with such prime test tools as Matlab and Simulink. "Simulink is wonderful for laying out your control system and getting the data acquisition set up," explained Faulring. "Then, when you want to exercise your model, such as running a motor at different speeds or loads, it is easy to take Matlab and write an M-file script to do that. It is also easy for the Simulink framework to set up parameters in your code that you can change in real time. And you don't need a GUI. It's easy to make tunable parameters on the fly while you are doing debugging."
Moving forward, Faulring and the rest of the Kinea team will have plenty of opportunity to use such tools. Kinea expects to be involved in a new DARPA contract to refine some of technologies pioneered in Revolutionizing Prosthetics 2009. The company also has received another DARPA contract to develop a robotic haptic hand to grasp and manipulate objects for potential use in hazardous environments, such as explosive ordinance disposal.
Kinea president Michael Peshkin foresees plenty of opportunity for human interactive mechatronics in applications ranging from academic research to industrial equipment to assistive devices for surgery. "Traditionally, people have viewed the robot as something to steer clear of," he explained. "What is new here is that humans and robots are working together hand in claw."
No related content found.
- 0 rated items found.
Datasheets.com Electronic Parts & Inventory Search
185 million searchable parts
- Part Number
- Description
- Inventory
- Products
- Manufacturers
View All Webcasts
