Underactuated Robotics

Algorithms for Walking, Running, Swimming, Flying, and Manipulation

Note: These are working notes used for a course being taught at MIT. They will be updated throughout the Spring 2024 semester. Lecture videos are available on YouTube.

Previous Chapter

Table of contents

Next Chapter

An Optimization Playbook

This chapter is intended to provide a collection of tips and formulations that can make seemingly non-smooth constraints smooth, seemingly non-convex constraints convex, etc. (It's very much a work in progress...)

Matrices

max eigenvalue. min eigenvalue. rank (nuclear norm, ...)

Ellipsoids

Imagine an ellipsoid centered at the origin: $E = \{x | x^T S x \leq 1\}$ with $S=S^T\succ0$. The volume of this ellipse is proportional to $\left(\det S^{-1} \right)^\frac{1}{2}$. We know that $\log\det(S)$ is a concave function in the elements of $S$ Boyd04a.

maximizing volume.

minimizing volume.

volume of a semi-algebraic set. (via a contained ellipsoid)

Polytopes

Sadra's linear encodings of polytopic containment

Perspective functions

(Mixed-)Integer Programming

Bilinear Matrix Inequalities (BMIs)

PQ trick from LQR Rank minimization

Geometry (SE(3), Penetration, and Contact)

Hongkai's smooth non-penetration constraints for polygons "Soft" absolute value: $|x|_epsilon \approx \sqrt{x^2 + \epsilon}$. import numpy as np import matplotlib.pyplot as plt x = np.linspace(-2.,2.,101) plt.plot(x, np.sqrt(np.power(x,2) + 0.01))

Barrier function for friction cones in Grandia 2019, Feedback MPC for Torque-Controlled Legged Robots. Linearization of the rotation matrix in Real-time Model Predictive Control for Versatile Dynamic Motions in Quadrupedal Robots Briat15 chapter 1 has some good modeling tricks for LPV systems

References

Stephen Boyd and Lieven Vandenberghe, "Convex Optimization", Cambridge University Press , 2004.

Previous Chapter

Table of contents

Next Chapter