EN4062 / ENT794 Advanced Robotics - Coursework 2
In thispart ofthe coursework, we will explore more on the SLIP mechanisms and how the optimization of legged locomotion is done. In Lecture 4, we explored the locomotion of a two-legged mass-spring (SLIP) model. A MATLAB toolbox (Bayesian Optimisation of SLIP.zip) is provided that is developed by Cambridge University. Download it to your desktop and unzip it. Open the folder in your Matlab program. Explore the code and read the “Readme” file to understand how the code works.
A- Run Results.m function and explain the meaning of each figure plotted on your screen. In your opinion, what is the best result of locomotion and why? Is there a relation between the fitness function and parameters in colourmap values of [α1 α2 k]? Why and how are these parameters related?
B- Let’s get deeper into how the optimization of locomotion behaves. Use BO.m file and run the optimisation for the following three initial condition cases:
A) Walking [x y x' y'] = [0 0.98 1.3 0],
B) Skipping [x y x' y'] = [0 0.95 1.6 0],
C) Running [x y x' y'] = [0 0.95 5 0]
Explain how the optimization work by changing the initial conditions for each case. What do you think is the reason for only changing initial conditions, the complete motion changes its behaviour and convergence (limit cycles) in the phase plot? In which case, is there a failing unstable phase plot? Show your results and discuss (no need to include the animation motion).
C- Find and change the values of touchdown angle and stiffness in the Bayesian Optimization code and make the SLIP model able to run (you can change initial conditions as well), but this time, the SLIP model during SSP keeps the vertical height around 5.5 m and stay stable at least 3 or more motion cycles without failing (can be confirmed by the phase plot). Explain how you designed this strategy of jumping and running. Describe your logic behind it.
Note 1: Please attach the values or codes for changed parts as an Appendix to this question and you should add snapshots of the SLIP model animation motion in your report body to present the result. The more successful you make the motion, the higher your mark for this Task 2-C will be.
Note 2: You can get an extra mark if you find any unusual and crazy motion patterns (different from what you observed in previous sections) with training certain values. The sky is the limit!
References
[1] Raibert MH, Brown Jr HB, Chepponis M. Experiments in balance with a 3D one-legged hopping machine. The International Journal of Robotics Research. 1984 Jun;3(2):75-92.
[2] van Dalen, S. "A linear inverted pendulum walk implemented on TUlip." PhD diss., Master’s thesis, Eindhoven University of Technology, 2012. D&C, 2012.
[3] Vukobratović M, Borovac B. Zero-moment point—thirty five years of its life. International journal of humanoid robotics. 2004 Mar;1(01):157-73.