ROB 501: Mathematics for Robotics
- Erogato da: University of Michigan (Michigan Robotics)
- Docente: Jessy Grizzle
- Livello: advanced
- Durata: 26 lectures (full semester, self-paced)
- Categoria: Mathematics
- Lingua: English
- Prezzo: Gratuito
- Tag: mathematics, kalman-filter, optimization, linear-algebra, least-squares, estimation
Il corso di matematica applicata di livello graduate della University of Michigan per ingegneri robotici, tenuto da Jessy Grizzle (Director of Michigan Robotics) e pubblicato gratuitamente per intero: 26 videolezioni, un libro di testo aperto, appunti delle lezioni, esercitazioni con soluzioni, dieci compiti a casa e due esami. Costruisce le basi matematiche della robotica moderna, dagli spazi vettoriali e dai minimi quadrati ai filtri di Kalman, alla convessità e all'ottimizzazione.
Perché seguirlo
This is the mathematical bedrock course of a top robotics graduate program, released in full for free. Kalman filtering, least squares, convex optimization — the machinery inside every perception, estimation and control stack — taught rigorously by Jessy Grizzle, one of the field's most respected educators and a legged-robotics pioneer. Most free robotics content teaches tools; this teaches the mathematics that makes the tools make sense.
Cosa imparerai
- Vector spaces, subspaces, bases and linear operators — abstract linear algebra done properly
- Projection theorem, Gram-Schmidt and least squares, through to recursive least squares
- The Kalman filter and its underlying probabilistic concepts, derived rather than recited
- Matrix factorizations (QR, Schur complement, positive semi-definite structure)
- Norms, convergent sequences, contraction mappings and Newton-Raphson
- Convexity, linear and quadratic programming for optimization in robotics
Programma
A full-semester graduate course (Fall 2018 offering) in 26 recorded lectures. It opens with proof techniques and abstract linear algebra, builds through inner product spaces, projection and least squares to recursive estimation and the Kalman filter, then covers matrix factorizations and closes with nonlinear optimization: Newton-Raphson, convergence, convexity, and linear and quadratic programming. Homework (10 sets), recitations with answers, two exams and an open textbook accompany the videos.
Argomenti principali
- Proofs, induction and contradiction as working tools
- Abstract linear algebra and inner product spaces
- Least squares: normal equations, recursive form, BLUE and minimum-variance estimation
- Kalman filter foundations
- QR factorization and numerical linear algebra
- Nonlinear optimization: Newton-Raphson, convergence, convexity, LPs and QPs
Prerequisiti
- Basic matrix algebra (multiplication, inversion, rank, eigenvectors)
- Elementary probability: means, variances, conditional probability
- Vector calculus (gradients, Lagrange multipliers) and basic MATLAB
A chi si rivolge
- Graduate students and advanced undergraduates entering robotics
- Engineers who use Kalman filters or optimizers and want to actually understand them
- Self-learners building a rigorous foundation before SLAM, control or learning courses
- Anyone following Michigan Robotics' free curriculum