2024-04-18 ExoBerlin Part I

About

So on the 16th and 17th of April 2024, the ExoBerlin conference took place in, well, Berlin. I attended the event and summarized what I learned from the exoskeleton-specific talks of professors, CEOs, and professionals of 9 different countries on this page.

Link

https://exo-berlin.de/

License

cc by 4.0

Credits

Author: Raphael Luif

Takeaway

  • Sensorless Motion Tracking technology is booming and making it into clinical applications
  • Parallel elasticity can greatly reduce motor requirements, energy consumption, and system responsiveness
  • The parallel elastic element can be optimized to best improve a certain aspect of the motor (peak torque, power usage, etc.)
  • An IMU can be used to counteract oscillations that may occur in control systems that use non-collocated force sensors with high sensitivity.

The presentations

Markerless gait analysis system for clinics

Syonko Galasso, a PhD Student from Lunex in Luxembourg talked about the need for and the advantages of markerless tracking and presented the MEMENTO Project. The MEMENTO Project is a three-year funded PhD project, which started in 2022 and aims to develop a markerless solution able to provide clinically relevant parameters quickly and inexpensively.

They used a 25-keypoint body model to determine the joint angles and other biometry.

The system was validated by comparing the data captured by 7 IMUs to the markerless one. Comparing the respective data showed that the markerless tracking reached 80% for compatibility and more than 70% for correlation. 

The project is not done yet, however the results seem promising.

Syonko mentioned that the system will be validated again with a marker-based system after this phase.

Notes:

They used OpenPose

https://github.com/CMU-Perceptual-Computing-Lab/openpose

Link

https://lunex.lu/

Advanced actuation and control strategies for affordable hybrid exoskeleton.

Andrea Calanca from the Altair robotics lab in the università di verona, Italy talked about how to reduce motor requirements for exoskeletons and improve system transparency. 

At first, he emphasized how important it is, no matter how advanced the assistive strategies of the exoskeleton if the low-level force controller is not fine-tuned for the job, the exoskeleton will not be of much help to the human user.  In this case, his experience is that starting at the low level, getting the force controller right is the way to go.  

In force control, you have two systems: The robot and the human. The dynamics of humans are very dynamic and oftentimes very hard to interpret and predict. This can lead to unwanted disturbances in the output force of the robot. 

Mr. Calanca found that if you have a low motor inertia you can reduce the sensitivity to external disturbances, aka the human.

Generally speaking about exoskeletons that scale the lower extremeties, most models use big oversized motors. With bigger motors come higher costs, higher weight and reduced ergonomics. To reduce the motor size and the motor inertia one can invest in a more power dense motor model, which in most cases is very expensive. But a more cost-effective solution would be to just add a spring in parallel.

A parallel elasticity in the system can help to reduce torque requirements which in turn helps to use smaller and less expensive motors. 

Next up Mr. Calanca introduced his exoskeleton design based on a parallelogram mechanism. A similar design can be seen in the Agadexo Shoulder for example. 

The above graph describes the systematic structure of a parallel mechanism shoulder exoskeleton. 1) Shoulder part, 2) elbow part 3) springs.

Figure taken from (A. Calanca, E. Dimo, E Palazzi, L. Luzi “Enhancing Force Controllability by Mechanics in Exoskeleton Design”) The proposed exoskeleton design concept. The parallelogram structure avoids torque propagation and passive gravity compensation. 

The above-seen design allows for easy implementation of gravity compensation and very low motor torque requirements. Furthermore, it interrupts the torque propagation from the shoulder to the forearm segment. 

With this new design, the required torque for a specific task could be reduced from the original 14 Nm to only 1 Nm. Further validation proved the advantages of the low motor inertia. Smaller motors are faster when it comes to changing directions, therefore the system becomes more resistant to human disturbances. 

So this exoskeleton got advantages over traditional exoskeleton designs (One big motor per human joint axis) by changing the mechanical structure and by adjusting the motor size. 

Next up Mr. Calanca spoke about how to optimize the spring which is offloading the motor. Again the goal is to reduce motor size by adding a spring in parallel with the motor. The motor specifications were presented in a radar chart. By introducing the spring certain characteristics (max. required torque, average squared torque, max power, etc.) could be optimized. So one should optimize a spring that reduces one of the motor specifications. One possible expansion of this exokseleton architecture is to make the elastic element adjustable. One way of doing it is to make it slide up and down on the shoulder part to change it´s angle attachment. You can see that working in the video on Agadexo shoulder´s website.

Inertia Transparency

Mr. Calanca showed a video in his presentation to us where the exoskeleton tried to move out of the users way to maintain a constant force at its force sensor. This force sensor might be located at one of the exos cuffs in the final version. In the video we saw that the system was very sensible, the exoskeleton moved out of the way even when the force sensor was only touched by a paper towel. 

This “constantly moving out of the way” of the exo is described as inertia masking or inertia transparency. So by using a force sensor, actuators and a control system the mechanical system friction and inertia can be virtually removed. This allows the user to move with the exoskeleton without feeling any kind of resistance even when the exo is heavy.   

Transparency, open issues

Non-collocation describes when the sensor and the motor are not mounted on the same rigid body. Any industrial robot has its torque sensor directly on the motor, this is crucial for the stability of the control system. Otherwise, the system usually starts to oscillate. Mr. Calanca showed us a video with the same inertia transparency control but this time instead of using a paper towel, he used a piece of steel to make the exoskeleton react. The exoskeleton started to vibrate instead of smoothly lifting as before. 

“As force control gains are raised to improve performance and transparency vibrations occurred(due to non-collocation)” (from one of the slides in his presentation)

So this high gain system, which in this context translates to high sensibility in combination with a non-collocated sensor is prone to vibration when in contact with a stiff material. 

A solution to this problem is to add an accelerometer, which detects when oscillations occur. This way vibrations can be compensated for. Mr. Calanca showed us in the next video that this approach indeed works and the actuators movements remained smooth even when the force sensor was interfaced with something very stfff. 

Mr Calanca mentioned that there he will be organizing an industrial exoskeleton workshop at CASE 2024.

References

  • A. Calanca, E . Dimo “Advantages of Acceleration Feedback on Force Control Performance and Transparency” in progress…
  • A. Calanca, E. Dimo, E Palazzi, L. Luzi “Enhancing Force Controllability by Mechanics in Exoskeleton Design”
  • E. Palazzi et al, “An Affordable Upper-Limb Exoskeleton Concept for Rehabilitation Applications” 
  • A. Calanca , S. Toxiri, D. Costanzi, E. Sartori, R. Vicario, T. Poliero, C. Di. Natali, D. G. Caldwell, P. Fiorini, J. Ortiz “Actuation Selection for Assistive Exoskeletons: Matching Capabilities to Task Requirements “
  • S. Toxiri, A. Calanca, P. Fiorini, and D. G. Caldwell “A Parallel- Elastic Actuator for a Torque-Controlled Back-Support Exoskeleton”