1. Hardware development on quadruped robots FINAL presentation
Welcome to my final presentation of my semester project about hardware development on quadruped robots.
Hardware development on quadruped robots FINAL presentation
Andreas GASSNER
2nd Semester Master Microengineering
Professor: Auke Jan IJSPEERT
Supervisors: Alexander SPRÖWITZ
Rico MÖCKEL
November 11, 2018

2. Here you can see the outline of my presentation.
I will start with explaining the motivation of this project, then, I will give a short introduction of the hardware of the oncilla project. Further on i will present my thre main tasks of this semester. I will explain them in detail and each part, the RC servo motor debugging and the walkway will be finished by a conclusion.
For your orientation i added the outline at the left side of each slide.
Outline
Motivation
Oncilla project - Hardware
Tasks of this semester project
RC servo motor debugging
Layout
Implemented CPG
Tests
Analysis of the results
Conclusion
Walkway
Walkway / Kistler force plate
Different configurations
Variants / Steel plates
Final design
11/11/2018
Andreas Gassner

3. The motivation of this project is, as it is described in the review of Professor Ijspeert, the interaction between robotics and neurobiology including not only robots benefiting of locomotion models but also the validation of biological models on robots. To achieve this goal two quadruped robots are build. This allows us to test different CPGs behaviour, gait transistions, leg ratios and feedback implementation not only in simulations but also on real robots.
Motivation
Motivation
Oncilla project
Tasks of the project
RC servo motor debugging
Layout
Implemented CPG
Tests
Analysis
Conclusion
Experimental setup
Walkway / Kistler
Diff. configurations
Variants/Steel plates
Final design
1 Central pattern generators for locomotion control in animals and robots: a review, IJSPEERT, Auke Jan, in Neural Networks, Vol 21/4, 2008
11/11/2018
Andreas Gassner

4. Oncilla project - Hardware
The oncilla project where Alexander Spröwitz is working on is part of the european amarsi project. In the Biorob 2 quadruped robot will be built, a small quadruped robot and a large oncilla robot. The small robot is based on the Cheetah platform. It serves as a quick test platform and has 2 RC servo motors per leg. The large oncilla has 2 bushless dc motors and 1 RC servo motor per leg. Finally a Central Pattern Generator with sensor feedback will be implemented on both robots. The simulations of the robots are then testet on a walkway.
Oncilla project - Hardware
Motivation
Oncilla project
Tasks of the project
RC servo motor debugging
Layout
Implemented CPG
Tests
Analysis
Conclusion
Experimental setup
Walkway / Kistler
Diff. configurations
Variants/Steel plates
Final design
2 quadruped robots
Small quadruped robot
Based on the Cheetah platform 1
Quick test platform
2 RC servo motors per leg
Large Oncilla
2 BLDC Motors per leg
1 RC servo motor per leg
CPG with sensor feedback
Walkway for testing and validation of simulations 2 3
1 Cheetah - compliant quadruped robot, RUTHISHAUSER Simon, 2008
2 Image from:
3 Image provided by Alexander Spröwitz
11/11/2018
Andreas Gassner

5. tasks of this semester project
My work is a subproject of the oncilla project. My three main tasks were the implementation and characterization of a CPG first in Matlab and then in C#. Secondly the debugging and initialization of a off-the-shelf RC servo motor control board. Where the CPG has then run through. Finally a completly different task but also linked to the project is the dseign of a walkway not only for quadruped but also for biped robots or the Amphibot and the implementation of the kistler force plate. In the next few minutes i will detail those tasks, where the first two tasks are strongly linked together.
tasks of this semester project
Implementation of a CPG
Simulation in Matlab
Transferred to C#
Debug and initialize RC servo motor control board
Pololu Mini Maestro 24
Run CPG through board
Design of a walkway
Quadruped, biped robots & AmphiBot
Kistler force plate
Motivation
Oncilla project
Tasks of the project
RC servo motor debugging
Layout
Implemented CPG
Tests
Analysis
Conclusion
Experimental setup
Walkway / Kistler
Diff. configurations
Variants/Steel plates
Final design 1 2
11/11/2018
1 Online trajectory generation in an amphibious snake robot using a lamprey-like central pattern generator model, IJSPEERT, Auke Jan and CRESPI, Alessandro, 2008 2
Andreas Gassner

6. RC servo motor debugging
So lets start with the rc servo motor debugging. The control board which was provided was the Pololu Mini Maestro 24.
RC servo motor debugging
Motivation
Oncilla project
Tasks of the project
RC servo motor debugging
Layout
Implemented CPG
Tests
Analysis
Conclusion
Experimental setup
Walkway / Kistler
Diff. configurations
Variants/Steel plates
Final design
Pololu Mini Maestro 24
11/11/2018
Andreas Gassner

7. This picture shows the layout which I used during my semester project with the frequency or tim limitations. Firstly a computer was used to calculate the cpg position which then was transfered throug the USB cable to the Mini Maestro 24. This control boards generates the PWM for all the rc servo motors which are placed on the robot. To do a more stable system and also to get the limitations of the system and to analyze the correct implementation, i tried to get the motors position back to the computer. Unfortunately the Mini Maestro does not provide this solution. So we had to find another solution which i will talk later.
Layout
Motivation
Oncilla project
Tasks of the project
RC servo motor debugging
Layout
Implemented CPG
Tests
Analysis
Conclusion
Experimental setup
Walkway / Kistler
Diff. configurations
Variants/Steel plates
Final design
No RC servo motor position feedback possible
Analysis of the correct implementation
Analysis of the amplitude / frequency
11/11/2018
1 Image taken from: ttp://
2 Image taken from:
Andreas Gassner

8. RC servo motor position feedback
This picture shows the layout which I used during my semester project with the frequency or tim limitations. Firstly a computer was used to calculate the cpg position which then was transfered throug the USB cable to the Mini Maestro 24. This control boards generates the PWM for all the rc servo motors which are placed on the robot. To do a more stable system and also to get the limitations of the system and to analyze the correct implementation, i tried to get the motors position back to the computer. Unfortunately the Mini Maestro does not provide this solution. So we had to find another solution which i will talk later.
Layout
Motivation
Oncilla project
Tasks of the project
RC servo motor debugging
Layout
Implemented CPG
Tests
Analysis
Conclusion
Experimental setup
Walkway / Kistler
Diff. configurations
Variants/Steel plates
Final design
RC servo motor position feedback
No RC servo motor position feedback possible
Analysis of the correct implementation
Analysis of the amplitude / frequency
11/11/2018
1 Image taken from: ttp://
2 Image taken from:
Andreas Gassner

9. RC servo motor Debugging
For the analysis i used this CPG with frequencies going from 1.5 Hz to 2.0 Hz. Plus-minus pi over four was chosen as an amplitude.
RC servo motor Debugging
Motivation
Oncilla project
Tasks of the project
RC servo motor debugging
Layout
Implemented CPG
Tests
Analysis
Conclusion
Experimental setup
Walkway / Kistler
Diff. configurations
Variants/Steel plates
Final design
Program sends CPG calculated position to control board
Test with load (500g), without load
Frequency1.5 Hz up to 2.0 Hz (step 0.25 Hz)
ai: 5 & ωij: 4
Number of oscillators: 8 (1 tracked)
Amplitude: ±π/4
Phase shift: 0
Implemented CPG 1:
𝜃 𝑖 =𝟐 𝝅 𝝂 𝒊 + 𝒋 𝝎 𝒊𝒋 𝒔𝒊𝒏( 𝜽 𝒋 − 𝜽 𝒊 − 𝝓 𝒊𝒋 ) 𝒓 𝒊 = 𝒂 𝒊 ( 𝒂 𝒊 𝟒 𝑹 𝒊 − 𝒓 𝒊 )− 𝒓 𝒊 𝒙 𝒊 = 𝒓 𝒊 (𝟏+𝒔𝒊𝒏( 𝜽 𝒊 )
11/11/2018
1 From swimming to walking with a salamander robot driven by a spinal cord model, IJSPEERT, Auke Jan et al. , Science, 2007
Andreas Gassner

10. So a CPG calculated position with an amplitude of pi over four and different frequencies was sent to the servo motor. Because no feedback from the servo is possible, we had to do a video aided point tracking. While running the motors are filmed and a point is then tracked frame per frame using the Mtracking tool provided by Kostas. The points then can be analyzed with matlab. The video analyzed points are supposed to be the real points. The whole experience was once made without load and once with a one sided load of 500g.
Video point-tracking
Motivation
Oncilla project
Tasks of the project
RC servo motor debugging
Layout
Implemented CPG
Tests
Analysis
Conclusion
Experimental setup
Walkway / Kistler
Diff. configurations
Variants/Steel plates
Final design 1
11/11/2018
1 Mtracking tool provided by Konstantinos Karakasiliotis
Andreas Gassner

11. Analysis of the Results
To analyze the range i ploted the points in matlab compared to the sent values. I then set the range of the sent values to 100% and calculated the corresponding percentage of the video tracked range. As one can see until 1.5 Hz the RC servo motors can follow the imposed positions with and without load. Then we can see a difference. The maximal possible frequency for the RC servo motors would be 2.08 Hz but in this case we dont take into account the accelerations.
Analysis of the Results
Motivation
Oncilla project
Tasks of the project
RC servo motor debugging
Layout
Implemented CPG
Tests
Analysis
Conclusion
Experimental setup
Walkway / Kistler
Diff. configurations
Variants/Steel plates
Final design
100%
Percentage of the full range (π/2)
Max RC servo motor frequency: 2.08 Hz
11/11/2018
Andreas Gassner

12. Analysis of the Results
To analyze the range i ploted the points in matlab compared to the sent values. I then set the range of the sent values to 100% and calculated the corresponding percentage of the video tracked range. As one can see until 1.5 Hz the RC servo motors can follow the imposed positions with and without load. Then we can see a difference. The maximal possible frequency for the RC servo motors would be 2.08 Hz but in this case we dont take into account the accelerations.
Analysis of the Results
Motivation
Oncilla project
Tasks of the project
RC servo motor debugging
Layout
Implemented CPG
Tests
Analysis
Conclusion
Experimental setup
Walkway / Kistler
Diff. configurations
Variants/Steel plates
Final design
100%
Percentage of the full range (π/2)
Max RC servo motor frequency: 2.08 Hz
11/11/2018
Andreas Gassner

13. So summarize this part:
The motor position feedback is not possible. So for analyzing the position i made a video analysis. For the correct implementation I had to adjust the interruption time. A real time implementation is necessary. Until now they are made with the system times but other solutions exist.
We also saw, that the amplitude of the movement increases when imposing too high frequencies and when putting loads.
Conclusion
Motivation
Oncilla project
Tasks of the project
RC servo motor debugging
Layout
Implemented CPG
Tests
Analysis
Conclusion
Experimental setup
Walkway / Kistler
Diff. configurations
Variants/Steel plates
Final design
RC servo motor position feedback
Not possible
Interruption time
Real-time implementation necessary
System-timer usage
Real time operating system
Implement CPG on microcontroller
Use better ODE solver method (e.g. Runge-Kutta method)
Decreasing amplitude at the motor at high frequencies and with load
11/11/2018
Andreas Gassner

14. Now we will go to my second task which was the design of a walkway for robot locomotion analysis.
Motivation
Oncilla project
Tasks of the project
RC servo motor debugging
Layout
Implemented CPG
Tests
Analysis
Conclusion
Experimental setup
Walkway / Kistler
Diff. configurations
Variants/Steel plates
Final design
11/11/2018
Andreas Gassner

15. Walkway
So the walkway will be used for verifying gait characteristics for example ground reaction forces already simulated on the computer. So with this walkway we will be able to do the transition from computer simulation to real robots. There will also be a base to install a motion tracking system.
Verifying gait characteristics
Ground reaction forces
Computer simulation to real robots
Possibly: installation of a motion tracking system
Motivation
Oncilla project
Tasks of the project
RC servo motor debugging
Layout
Implemented CPG
Tests
Analysis
Conclusion
Experimental setup
Walkway / Kistler
Diff. configurations
Variants/Steel plates
Final design
11/11/2018
1 Exploiting Compliance with a Cat-sized Quadruped Robot for Trot Gait, TULEU, Alexandre et al.
Andreas Gassner

16. Two Kistler force plates
To do so, we the lab possesses two kistler force plates which you can see here.
Two Kistler force plates
Motivation
Oncilla project
Tasks of the project
RC servo motor debugging
Layout
Implemented CPG
Tests
Analysis
Conclusion
Experimental setup
Walkway / Kistler
Diff. configurations
Variants/Steel plates
Final design
11/11/2018
Portable multicomponent force plate for gait- and balance analysis in biomechanics
Andreas Gassner

17. Different configurations
While this walkway is used for all kind of robots the quadruped the biped robots and also the amphibots we have to make possible different configurations. But also to analyze different gaits. Some examples are shown here.
Different configurations
Motivation
Oncilla project
Tasks of the project
RC servo motor debugging
Layout
Implemented CPG
Tests
Analysis
Conclusion
Experimental setup
Walkway / Kistler
Diff. configurations
Variants/Steel plates
Final design
For AmphiBot, quadruped & biped robots
Different gaits
11/11/2018
Andreas Gassner

18. Force plate setup variants
We elaborated two variants, the variant alu and the variant steel. In the variant alu alimine frames are glued with epoxy or screwed to the lab floor. The variant steel uses steel plates to fix the force plates. Because of this three main advantages we choosed the variant steel.
Force plate setup variants
Variant alu
Variant steel
Motivation
Oncilla project
Tasks of the project
RC servo motor debugging
Layout
Implemented CPG
Tests
Analysis
Conclusion
Experimental setup
Walkway / Kistler
Diff. configurations
Variants/Steel plates
Final design
Advantages
Low installation costs
Frames already present in the lab
Disadvantages
Height not adaptable
Destruction of the lab floor
Cost estimation
SFr
Advantages
Adjustable height and declination
Avoid vibrations
No destruction of the lab floor
Flexibility
Disadvantages
Construction necessary
Expensive
Inner stress possible
Cost estimation
SFr
11/11/2018
Andreas Gassner

19. The base for the force plates have to be at least 10 times as heavy as the force plates weicht so around 60kg. The steelplates are made that the different configurations you can see here are possible. Machine foots should avoid lab floor vibrations. The steel plates are providing high inertia.
In this configuration while two force plates are on one side two steel plates have to be put together to respect the instructions given in the manual.
Steel plates
Motivation
Oncilla project
Tasks of the project
RC servo motor debugging
Layout
Implemented CPG
Tests
Analysis
Conclusion
Experimental setup
Walkway / Kistler
Diff. configurations
Variants/Steel plates
Final design
High inertia
Manual: base weight > 10x force plate weight
11/11/2018
Andreas Gassner

20. The base for the force plates have to be at least 10 times as heavy as the force plates weicht so around 60kg. The steelplates are made that the different configurations you can see here are possible. Machine foots should avoid lab floor vibrations. The steel plates are providing high inertia.
In this configuration while two force plates are on one side two steel plates have to be put together to respect the instructions given in the manual.
Steel plates
Motivation
Oncilla project
Tasks of the project
RC servo motor debugging
Layout
Implemented CPG
Tests
Analysis
Conclusion
Experimental setup
Walkway / Kistler
Diff. configurations
Variants/Steel plates
Final design
High inertia
Manual: base weight > 10x force plate weight
11/11/2018
Andreas Gassner

21. Here you can see the final design
Here you can see the final design. Wooden plates will build the walkway, a wooden frame will give the neded height. The middle wooden plates can be changed depending on the chose configuration .
Final design
Motivation
Oncilla project
Tasks of the project
RC servo motor debugging
Layout
Implemented CPG
Tests
Analysis
Conclusion
Experimental setup
Walkway / Kistler
Diff. configurations
Variants/Steel plates
Final design
11/11/2018
Andreas Gassner

22. So for summarize:
A walkway was designed and two variants for foce plate fixation were proposed. The variant steel was recommended and construction files are provided. An order for all the material was prepared. A request for price and feasibility of the steel plates has been made.
conclusion
Motivation
Oncilla project
Tasks of the project
RC servo motor debugging
Layout
Implemented CPG
Tests
Analysis
Conclusion
Experimental setup
Walkway / Kistler
Diff. configurations
Variants/Steel plates
Final design
Walkway with embedded force plates
Two variants proposed
Variant steel recommended
Construction files provided
Order prepared
Request for price and feasibility
11/11/2018
Andreas Gassner

23. Thank you for your attention
Thank you for your attention. I’m sure you have some questions.
Questions?
Thanks to:
Professor Auke Jan Ijspeert
Supervisors Alexander Spröwitz & Rico Möckel
Jonathan Grizou, Rolando Rodas, Laurenz Notter

24. 11/11/2018
Andreas Gassner

25. So for finishing the small oncilla platform i have to resolve the timer problem i mentioned, then i can implement the CPG programmed in C# and the small oncilla should be ready for some first tests.
Concerning the large oncilla, firstly i have to make some reading on the RoBoard RB-110, which is the controller board of the large Oncilla. Then the brushless dc motors have to be implemented. For executing the tests on both Oncillas the experimental setup has to be created.
Future work
Order and build the walkway
Build the robot and implement the program
Tests on the platform using the experimental setup
Possibly installation of a motion tracking system
Implementation of a more complex gait able to produce a locomotion movement
Implementation of a real time solution
11/11/2018
Andreas Gassner

26. Effect of parameters ωij = 0 , ai = 100 ωij = 1 , ai = 100
11/11/2018
Andreas Gassner

27. Debugging Cropped sine waves (timer problem)
Maestro control center (min-max values)
Video point-tracking analysis
11/11/2018
Andreas Gassner