Introduction

Start the course in the hallway, not in the simulator. Point at a vacuum that docks itself, a warehouse shuttle, a quadcopter that holds station in wind. Ask: who is closing the loop? If the answer is “a person with a radio,” you do not yet have an autonomous robot. If the answer is “a program that still has to guess about a messy world,” you do.

Autonomy is not a binary badge. It is a budget: how much of the next second is decided on board, from sensors that lie a little, actuators that saturate, and a map that is already stale. The classical picture is sense, plan, act. Real robots blur the boxes. A reflex may fire before a planner finishes. A planner may wait on a perception stack that is still voting on whether that blob is a person. Your job in this masterclass is to keep the boxes honest — to know which computation belongs where, and what physical facts it is allowed to assume.

The book is organized as four parts for a reason. Mechanisms tell you what motions are even possible. Sensing and actuation tell you how energy and information enter and leave the machine. Computation turns those streams into decisions. Uncertainty is the tax you pay for living in a world you only measure. Students who skip ahead to “the SLAM chapter” without wheels, cameras, and a little statics can write a filter that looks correct and still drives into a wall. Students who only ever tighten screws never notice that their odometry is a random walk.

Treat the first week as a contract. We will not paste the textbook into the browser. We will not compile a PDF and host it. We will write in our own words, then send you back to the chapter. That is the license talking, and it is also good pedagogy: a companion that replaces the book teaches you to skip. A companion that argues with the book teaches you to read.

When you meet a new robot, ask four questions in order. What can it physically do? What can it measure? What algorithm is allowed to run in the time it has? How wrong can those measurements be before the algorithm becomes theater? If you can answer those, you already have the syllabus in your pocket. The rest of the lessons are those four questions, slowly, with equations.

A useful classroom move: pick one robot and refuse to change it for a week. A differential-drive base with a lidar is enough to touch kinematics, sensing, planning, and localization. A manipulator on a table is enough to touch forces, grasping, and inverse kinematics. Switching platforms every lecture makes every idea look harder than it is. Hold the machine fixed and let the math move.