The presentation

Working with an Open-Source Lower Limb Exoskeleton Platform

Dr. Robert Trott presents his work on a novel controller for lower limb exoskeletons for gait rehabilitation. Dr.Robert Trott comes from Australia and works with a small team of six specialists for the continuation of his PhD work. The goal is to create an intuitive control algorithm for exoskeletons for stroke patients to better aid the rehab process.

To start, Dr. Trott had to get his hands on a commercial lower limb exoskeleton that he could then start to modify. The ones listed on his slides were the eksobionicsNR, the Indego, Cyberdyne HAL, and the Fourier ExoMOTUS X2. The exoskeleton needed to be capable of fully carrying the user, so no crutches are needed. The second very important criteria was modifiability. Dr. Trott was in contact with the manufacturers of the exos, and unsuprisingly, most of them were not really happy with the fact that the exoskeleton would be modified. In the end, Dr. Trott and his team went with the Fourier ExoMOTUS X2.

About the ExoMOTUS X2

The exo weighs in at around 18kg and supports sitting, stepping, and walking of quadrupedal gait (bipedal + crutches). The exoskeleton is actuated around the hips and knees. Each joint can deliver up to 100Nm and has a maximum rotary velocity of 4.5 rad/s. It uses CANBus for communication and trajectory control. Essentially, the heavy thinking of the exoskeleton is done in the Main Motion Unit MMU (branded name, aka. CPU), which gets its input from the Hand Control Unit HCU and Inertia Measurement Unit IMU. The HCU has four buttons for the user to manually choose which mode (sit, stand, step, and walk) he/she needs. Connected to the CPU are the four motors with their respective controllers, and four Analog Sensing Units, or ASUs for short. The ASUs get sensor data of the ground reaction forces and cuff interaction forces and turn them into a CANBus signal.

Modifications

First of all, the MMU and HCU were replaced with a new controller to enable the team to make software modifications and take control of the exoskeleton. Furthermore, because only one limb of a stroke patient is affected, the intent is to create a unilateral device, meaning that the exoskeleton only extends over one leg. This way, the affected limb gets supported by the exo and the unaffected limb stays unconstrained. Dr. Trott mentioned,  “Undoing a couple of bolts and unplugging leg sensors was the extent of that.” Next up was to redesign the emergency stop, or E-Stop, and power supply. The X2 had a soft E-Stop meaning it was implemented in the software, relying on the MMU to deactivate the actuators as opposed to an actual switch that cuts the power. The power supply had to be redesigned to separate the high-voltage rail for the actuators and the logic voltage for the computers and sensors. This way, the computers could stay on and record data even after using the E-Stop.Lastly, a new case needed to be produced to cover the newly fitted electronics. 

Software

Dr. Trott mentioned CORC a couple of times. (From the respective github sites) CORC CANopen Robotic Controller is a free and open-source robotic development software stack. It was initiated at the University of Melbourne in partnership with Fourier Intelligence (The guys where they got the exoskeleton from). The project was initially developed to run on the ExoMotus X2 Exoskeleton powered by a Beaglebone Black (…).
Essentially, Mr. Trott explained that there are three levels to CORK:

• The CANBUS, the serial bus itself, enabling coimmunication between components.
• The robot level, where you define the actuators and components of the robot.
• And the application level, where most of the testing is done, by adjusting parameters.

Another term Dr. Trott mentioned is ROS. ROS or Robot Operating System is a open-source set of software libraries to help you build robot applications.

I am not going to go more in depth here; serial-bussed and software are topics of their own. However, it is cool to see that not only researchers use open-source software and hardware (to some extent) but also companies (to some extent). Additionally, it is very valuable to see what hardware and software commercial exoskeletons and researchers use as a starting point if you want to build your own exo.

Adjustments were made on the robotic level for the drives, like the acceleration curves, maximum velocities, and torque control commands. Additionally, the homing cycle has been reworked. (Pretty interesting that the axis have to home, meaning they do not use absolute encoders.) Custom applications were written to control the exoskeleton, among other things, like safeguards. Getting ROS and CORC to communicate with each other was a major challenge, Dr. Trott said. 

Applications

As previously mentioned, customs applications were written for the exoskeleton. Notable is that there was a gait application with position, velocity, and torque control, respectively, written. The position control resulted in a very jittery motion of the exoskeleton (probably because of the choice of motor). Additionally, programs have been written for jogging with velocity and torque control (I am not sure if that means actual jogging like in sports or jogging like a single-axis movement in the context of CNC machines). -Sit to Stand with Naive and CGA.-Implemented a controller. Future work includes a force transparency application so that the exo moves out of the way of the user’s movements. 

Future work.

Trott and the team successfully got the X2 to walk “after frying some components and putting it back together”. There are plans to get the modified X2 into a clinic and let stroke patients try it out. They also want to add one feature that I have not seen with commercial rehab exoskeletons: actuation for abduction and adduktion. I am wondering how that will impact a normal gait, or if it has the potential to help the patient keep his/her balance. Furthermore, Dr. Trott said that he wants to replace the ground reaction force sensors. Additionally, strain gauges on the exoskeleton frame, among other additional sensors, should be added in the future. 

Personalization

Dr. Trott mentioned that he is in direct contact with some of the patients and found out that different patients have different needs for their ankle orthoses. Depending on how the gait is altered by the injury or sickness, one patient might prefer a stiffer ankle orthosis or a more compliant, flexible one. I imagine that this kind of personalization can greatly increase the ergonomics of the patient wearing the exoskeleton and therefore increase the likelihood of a faster rehabilitation. I hope that this kind of personalization will be adopted by more exoskeleton producers. 

Thoughts

I really enjoyed Dr. Trott´s talk; he was not afraid to talk about the many times some component got fried or something went wrong. It seems like a lot of tinkering and try and error was involved, as it usually is with prototyping. The talk presented a way to validate a control algorithm for an exoskeleton. The result that was presented was an electronic hardware overhauled exoskeleton with working applications written for walking, among other tasks. With this setup, the control algorithm can be tested not only virtually but in the real world and can further be tested by stroke patients for quick feedback.