ME410
From CYPHYNETS
ME410: Special Topics in Control Theory / EE396: Special Topics in Signal Processing
Summer 2011: Probabilistic Methods in Robotics & Control 

Instructor
Dr Abubakr Muhammad, Visiting Assistant Professor of Mechanical Engineering
Email: muhammad.abubakr [at] kaust.edu.sa, abubakr [at] lums.edu.pk
Office: 3303, AlKhwarizmi Bldg (1), KAUST
Office Hours: Sat, Sun, Tues 4:45pm5:30pm (or by appointment)
Course Details
Official title: ME410/EE396. Topics in contemporary control theory & practice / Advanced topics in signal processing.
Semester: Summer, 2011
Credits: 3
Status: Grad elective for electrical engineering, mechanical engineering, computer science, applied mathematics majors
Prereqs: Undergraduate level signals and systems; linear algebra; basic probability; programming experience
Dates: 11/06/201102/08/2011
Time: 15:0016:40
Days: Sat, Sun, Tue
Classroom: 2120
Course Website: http://cyphynets.lums.edu.pk/index.php/ME410
Course Description
A researchmethods based course to study probabilistic methods in robotics and control system design with emphasis on field robotics, unmanned aerial and ground vehicles, planning algorithms, autonomous systems, telerobotics, Human Robot Interaction (HRI) and other related areas. The course prepares students to do independent work at the frontiers of robotics and control research.
Course Contents
Introduction to basic concepts in robotics, Review of basic probability, random signals and systems, classical parameter estimation (min variance, least squares, max likelihood), Noisy sensor and motion models in robotics, concepts of statespace, Kalman filtering, Extended Kalman filter (EKF), robot configuration spaces, motion planning problems, Simultaneous Localization and Mapping (SLAM), Bayesian filtering, MonteCarlo methods, particle filtering, SLAM variants, Unscented and Ensemble Kalman filters; practical issues in autonomous robotics; examples/exercises from control, signal processing.
Objectives
 Introduce fundamental principles in robot motion planning.
 Use of geometric and dynamical models.
 Using sensorbased information to determine robotâ€™s own state and of the world.
 Linear and nonlinear statistical estimation techniques applied to problems in robotics and dynamical systems.
 Introduce practical applications
This course is NOT about
 Mechatronics or robot building.
 Higherlevel perception and AI.
Similar courses
LUMS. http://cyphynets.lums.edu.pk/index.php/CMPE633 (my course last fall)
Stanford. http://robots.stanford.edu/cs22606/schedule.html
CMU. http://www.cs.cmu.edu/~thrun/16899/
Centro de Investigacion en Matematicas, Mexico. http://www.cimat.mx/~jbhayet/
Univ of Washington. http://www.cs.washington.edu/education/courses/cse571/07au/
Text book
The course will be taught from a combination of the following textbooks.
 Principles of Robot Motion by Choset et al. MODULE 1
 Optimal State Estimation: Kalman, H Infinity, and Nonlinear Approaches by Dan Simon. MODULE 2
 Probabilistic Robotics by Thrun et al. MODULE 3
Grading Scheme
 Class Participation: 5%
 Assignments: 20%
 Midterm Exam: 25%
 Project: 50%
 Proposal. 10%
 Report. 20%
 Presentation. 20%
 Code/demo. 50%
Video Recordings
The entire course has been recorded on video and available on KAUST media website http://mediasite.kaust.edu.sa/mediasite/Catalog/catalogs/default.aspx . The videos can only be accessed from within KAUST campus or remotely via KAUST VPN. Please navigate to KAUST Main Catalog>Academic>Physical Science & Engineering>ME420/EE396 or follow direct links to the videos given below.
Schedule
WEEK  TOPICS  READINGS/REFERENCES 

Week 1. June 11  MODULE 1. Introduction to Robotics
Lec 1. Introduction; robotics and autonomous systems; Video Lec 2. Robotic workspaces; planning algorithms; bug algorithms; Bug0 and Bug1 algorithms; Completeness of Bug1; upper bounds on Bug1; Bug2 algorithm; performance comparison; range sensors and mathematical description; Tangent Bug algorithm; implementation issues; wallfollowing behavior with range sensors; some practical sensors; Lec 3. Robot configuration spaces; forward and inverse kinematics; mapping workspace obstacles into configuration space obstacles; visualizing high dimensional configuration spaces; a first look at obstacles in SE(2); polygonal robots and obstacles; configuration spaces of linkages; visualizing tori; Video 
Choset CH 2, 3, Appendix F; L1 Slides (From CMU RI16735) 
Week 2. June 18  Lec 4. Lifting velocities in forward and inverse kinematics; Rigid body transformations; geometry and topology of SO(2), SE(2); Video
Lec 5. Geometry and topology of SO(3), SE(3); Euler angle parameterization; Review of linear algebra; matrix calculus; Video MODULE 2. Review of Linear Systems and Probability Lec 6. Review of linear system theory; statespace models; matrix exponential; continuous to discrete conversions; difference between state space and configuration space in robotics; Video  Simon CH 2, 3;
Jordan Canonical Form (Handout). Gimbal lock problem in Euler Angles. Video. 
Week 3. June 25  Lec 7. Lyapunov and exponential Stability for LTI systems; controllability; observability; rank tests; examples; Review of basic probability; Bayes rule; random variables; transformations of random variables; Video
Lec 8. Linear transformation of Gaussian RV; Graphical interpretations; NonGaussian distributions arising due to nonlinearities; mutliple random variables; independence, correlation and correlation; vector random variables; autocovariance and autocorrelation matrices; introduction to stochastic processes; Video MODULE 3. Linear Estimation and Kalman Filtering Lec 9. Stochastic processes; strict sense and wide sense stationarity; ergodicity; introduction to state estimation problem; a relook at the averaging formula; least squares estimation problem; batch linear estimation; modeling noise covariance; generalized least square;  Simon. CH3; 
Week 4. July 2.  Lec 10. Recussive least square estimation; unbiased least squares estimators; optimal gain calculation; covariance and state update equations;
Lec 11. Generalized recursive least squares; Alternate estimator forms; examples and applications; stochastic linear differential equations; process noise; propagation of mean and covariance of a noisy LTI system; Lyapunov equation for steady covariance; Video Lec 12. Student project proposal presentations; Linear estimation of noisy LTI systems; notation and problem setup; Video  Simon. CH3, 4; 
Week 5. July 9.  Lec 13. Kalman filter derivation; examples; tracking of slowly varying paramters; noisy robot motion models; [1]
Lec 14. Kalman Filtering simulations for estimating slowly varying paramters; discretization of process noise covariance from continuous white noise models; constant velocity model derivation; Extended Kalman Filter; Linearlizaton on nonlinear state evolution and measurement; EKF applied to Range and Bearing localization; Video MODULE 4. Bayesian filtering methods Lec 15. Bayesian filter; derivation; simple state estimation examples; histogram filtering; introduction to particle filtering; localization demos; Video  Simon CH5; Thrun CH2, 4; Choset CH8,9;
Slides (From Stanford CS226)

Week 6. July 16  Lec 16. Nonparamteric filtering; Derivation of particle filter; analysis of resampling and particle weight computation; Binary Bayes filter for static estimation; Video
Midterm Exam. Lec 17. Robot perception models; beam models; sensor noise models; EM algorithm for learning sensor noise pdf from real data; Video  Thrun CH4, 6; 
Week 7. July 23  Lec 18. Likelihood fields for sensor models; robot motion models; odometry model; velocity motion model. Video
Lec 19. Occupancy grid maps; derivation of binary Baye's filter; map updates; Introduction to SLAM problem; Video Lec 20. General SLAM problem setup; EKF SLAM; derivation of EKF SLAM algorithm; landmarktofeature correspondences; EFKSLAM with correspondence learning; Video  Thurn CH5, 6, 9, 10 
Week 8. July 30  Lec 21. Information vector/matrix parametrization of a multivariate Gaussian; Information filter and its equivalence to Kalman filter; GraphSLAM algorithm; derivation of GraphSLAM; nonlinear least square formulation; Factorization of information matrix and GraphSLAM speedup; GraphSLAM and EKFSLAM comparison; course windup. Video
Student project presentations.'  Thurn CH2, 11 
Student Projects
 WiFi GraphSLAM in 1D. Jose Roberto Ayala Solares. Ref Paper Final Report Presentation
 CommunicationAware Motion Planning. Doha Hamza. Ref Paper Final Report Presentation
 WiFiSLAM Using Gaussian Process Latent Variable Models. Imran Shafique Ansari. Ref Paper Final Report Presentation
 Robot Localization in a Known Shopping Street Environment. MustafizurRehman Khan. Ref Paper Final Report Presentation
 EKF SLAM Using Prior Information. Samuel Castaneda. Ref Paper Final Report Presentation