A robot at a family’s kitchen sink must choose the right wash programme for every steel plate, tumbler, bowl, kadai, and delicate non-stick tawa that comes its way. This chapter puts the whole Chapter 2 toolkit to work on that one decision: modeling the costs of mistakes, fusing sensors, and learning sensor models from data.
| Section | Question | What you will learn |
|---|---|---|
| 3.1 Modeling the Wash Station | What is the robot deciding? | states, programmes, and a cost matrix for under-cleaning, over-cleaning, and damage; the no-sensor decision; the oracle and the value of perfect information |
| 3.2 Sensing Vessels | What do the sensors say? | camera, scale, and grease sensor models; naive Bayes fusion; why accuracy is not the same as cost |
| 3.3 Learning from Examples | Where do sensor models come from? | maximum-likelihood estimates; Laplace smoothing; learning curves and confusion matrices; more data or a better sensor? |
| 3.4 Summary and Exercises | key ideas, equations, review questions, numerical problems with answers, mini-projects |
This chapter builds directly on Section 2.1 (the vessel PMF) and Section 2.4 (expected cost).