RSS
Robotics: Science and Systems (R:SS) Course Webpage
This course will be a Masters degree level introduction to several core
areas in robotics: kinematics, dynamics and control; motion planning;
state estimation, localization and mapping; visual geometry, recognition of textured objects, shape matching and object categorization.
Lectures on these topics will be complemented by a large practical that
exercises knowledge of a cross section of these techniques in the
construction of an integrated robot in the lab, motivated by a task such
as robot navigation. Also, in addition to lectures on algorithms and
lab sessions, we expect that there will be several lecture hours
dedicated to discussion of implementation issues - how to go from the
equations to code.
The aim of the course is to present a unified view of the field,
culminating in a practical involving the development of an integrated
robotic system that actually embodies key elements of the major
algorithmic techniques. NOTE: This is a 20 pt course, as opposed to the standard 10 pt courses since this covers two introductory topics: robotics and vision and a practical element.
When and Where?
When: 9:00 - 10:50 (with 10 min. break) on Mondays and Thursdays.
Where: Monday = LT270 OC (Old College), Thursday = F.21 7GSQ (7 George Square)
First Lecture: 20 Sep (Thu) 9:00-10:50 @ F.21 7GSQ (7 George Square)
Summary of intended learning outcomes
- Model the motion of robotic systems in terms of kinematics and dynamics.
- Analyse and evaluate a few major techniques for feedback control, motion planning and computer vision as applied to robotics.
- Translate a subset of standard algorithms for motion planning, localization and computer vision into practical implementations.
- Implement and evaluate a working, full robotic system involving elements of control, planning, localization and vision.
Assessment
Written Examination 50
Assessed Practicals 40
Assessed Assignments 10
Lecture plan
Lecture time: 9:00 - 10:50 (with 10 min. break) on Mondays and Thursdays.
| Week |
Date | Lecture notes |
Lecturer |
Lecture topic |
Milestones |
|---|---|---|---|---|---|
| 1 |
20-Sep-2012 | Notes.1 |
Sethu Vijayakumar |
Introduction; Notations, Transformations, Rotations (1h15mim), Primer for the Practicals (30min) |
|
| 2 |
24-Sep-2012 | Notes. 2 Intro Notes. 2 Image Formation |
Vittorio Ferrari |
Image acquisition: basic world-to-image geometry and color spaces (1h); Two-view geometry: setting, notion of point correspondences, transformation classes for planar objects: similarity, affine, homography (1h) | |
| 2 |
27-Sep-2012 | Notes.3 Kinematics |
Sethu Vijayakumar | Kinematic (Forward, Inverse), Jacobian, Operational Space, Null Space, Optimality Principles (2h) |
|
| 3 |
1-Oct-2012 | Notes. 4 Path Planning | Subramanian Ramamoorthy |
Introduction to path planning methods |
Practicals (Wed.): Robot is able to move around. |
| 3 |
4-Oct-2012 | Notes. 5 Two View Geometry | Vittorio Ferrari | Two-view geometry: fundamental matrix (properties and estimation), invariance classes, invariants for planar configurations of points and lines |
Practicals (Thu.): Robot is able to move around. |
| 4 |
8-Oct-2012 | Notes. 6 Motion Planning | Subramanian Ramamoorthy | Sampling based path/motion planning |
Practicals (Wed.): Obstacle avoidance. |
| 4 |
11-Oct-2012 | Notes. 7 State Estimation | Subramanian Ramamoorthy | State estimation | Practicals (Thu.): Obstacle avoidance. |
| 5 |
15-Oct-2012 | Notes. 8 Interest Points | Vittorio Ferrari | Implementation issues for homography and fundamental matrix estimation (1h); Interest points and regions: general concept, plain Harris, scale-invariant Harris (1h) | |
| 5 |
18-Oct-2012 | Notes. 9 Feature Matching | Vittorio Ferrari | Interest points and regions: affine-invariant IBR and MSER (1h); implementation issues (1h) | Assignment 1 |
| 6 |
22-Oct-2012 | Notes.3 Kinematics (cont'd) Notes.10 Dynamics |
Sethu Vijayakumar | Kinematic and multi-objective motion planning (1h), Dynamics: Point mass, PID, Newton Euler, Joint Space, Optimal Operational Space Control, Non-holonomic sytems (1h) |
|
| 6 |
25-Oct-2012 | Notes. 11 Affine features Notes. 11 Specific object recognition |
Vittorio Ferrari | Specific object recognition: global descriptors, interest point/region descriptors (SIFT, moments), matching interest points/regions, filtering mismatches with geometric consistency (local consistency tests, global consistency tests with RANSAC) |
|
| 7 |
29-Oct-2012 | Notes.10 Dynamics (cont'd) Notes.12 Control SOC Additional Notes |
Sethu Vijayakumar | Dynamics (cont'd) (1h); Control: Intro to Optimal Control, HJB equations, LQR (1h) |
Practicals (Wed.): Resource identification. |
| 7 |
1-Nov-2012 | Notes. 13 SLAM | Subramanian Ramamoorthy | Localization and Mapping | Practicals (Thu.): Resource identification. |
| 8 |
5-Nov-2012 | Notes. 14 Edge detection | Vittorio Ferrari | Specific object recognition: correspondence expansion, how to do it very fast for large-scale object/image retrieval (1h); implementation issues (1h); |
Assignment due Practicals (Wed.): Visual servoing. |
| 8 |
8-Nov-2012 | Notes. 15 Image segmentation | Vittorio Ferrari | Edge detection and segmentation: simple thresholding, convolutions, canny, graph-cut, grab-cut |
Practicals (Thu.): Visual servoing. |
| 9 |
12-Nov-2012 | Notes. 16 Motion Synthesis | Subramanian Ramamoorthy | Motion synthesis in dynamic environments | Practicals (Wed.): Homing. |
| 9 |
15-Nov-2012 | Guest Lecture Info Sheet Notes. 17 Compliant Motion Control |
Wyatt Newman | Guest Lecture by Wyatt Newman (Venue: IF 4.31) Impedance and Force Control |
Practicals (Thu.): Homing. |
| 9 |
16-Nov-2012 11:00-12:00 and 13:00-14:00 |
Wyatt Newman |
Guest Lecture: Wyatt Newman (Non-Examinable) (Venue: IF 4.31) Compliant Motion Control: Applications and Implementations |
|
|
| 10 |
19-Nov-2012 | Notes. 18 Shape matching | Vittorio Ferrari | Shape matching: global descriptors, shape signatures, shape contexts, etc. |
|
| 10 |
22-Nov-2012 | Notes. 19 Object categorization | Vittorio Ferrari | Object categorization taster: problem definition and challenges, two simple models (generalized hough transforms, sliding-windows), learning parameters from training data, part-based models, the need for weak supervision. THIS LECTURE WILL NOT BE PART OF THE EXAM. |
|
| 11 |
26-Nov-2012 | Final Demo: Practice | |||
| 11 |
29-Nov-2012 | Final Practical Demo / Competition | Competition |
Recommended Texts
- Siciliano, B., Sciavicco, L., Villani, L., Oriolo, G., Robotics: Modelling, Planning and Control
- H. Choset, K.M. Lynch, S. Hutchinson, G. Kantor, Principles of Robot Motion: Theory, Algorithms, and Implementations.
- S. Thrun, W. Burgard and D. Fox, Probabilistic Robotics.
- D.A. Forsyth, J. Ponce, Computer Vision: A Modern Approach.
