University
Robotics as a degree course teaches it: kinematics, odometry, filtering, control, estimation, mapping, planning and vision, written as programs that run on a robot which does not know where it is.
Start in the simulatorU1. The robot as a system
What a robot program is really doing: a loop that senses, decides and acts, on hardware that answers late, imperfectly, and without…
7 lessons
U2. Kinematics and frames
Body frame and world frame, the rotation matrix between them, and the inverse kinematics that turn a velocity you want in the world into…
7 lessons
U3. Odometry and drift
Writing your own dead reckoning from the flow sensor and the gyro, watching its error grow, calibrating the two largest causes out, and…
7 lessons
U4. Noise and filtering
A reading is a number plus a distribution. Measuring the distribution, averaging it down, filtering in real time, throwing out the…
7 lessons
U5. Feedback control
Proportional, integral and derivative: what each term is for, how to tell from a chart which one you need, how to tune them on a real…
7 lessons
U6. State estimation
Two imperfect sources of the same number, and how to combine them: the complementary filter, then predict and correct, then the Kalman…
7 lessons
U7. Localisation
Working out where the robot is when it does not already know: a belief made of samples, moved by the motion model, weighted by the…
7 lessons
U8. Mapping
Turning range readings into a map: a grid of cells, an inverse sensor model that marks a ray free and its end occupied, log odds so the…
7 lessons
U9. Planning
Finding a route before driving it: the configuration space, a grid as a graph, breadth first, Dijkstra, A* and what makes a heuristic…
7 lessons
U10. Following a trajectory
Turning a plan into motion: a path is geometry and a trajectory is geometry with a clock, and between them sit velocity profiles,…
7 lessons
U11. Vision
The camera as a measuring instrument: what a pixel really tells you, what calibration measures, how range from apparent size falls apart…
7 lessons
U12. Learning, and the capstone
Working a behaviour out from data instead of writing it: fitting a model, judging it on data it has not seen, searching for a policy on…
7 lessons