Lecture 1 – January 8th, 2018

You will learn about stereo vision: how to estimate depth.

Objectives

• To understand the typical sensing modalities used in modern mobile robots: different types of sensors, how they work, how the data is processed, etc…
• To become familiar with different approaches of localisation and navigation in mobile robotics: where you are, and how to get to your destination.
• To learn how to design, implement and evaluate robotics algorithms in real-world application scenarios: in the labs you’ll work with Lego NXT robots.

Fundamental Questions

• Where am I? $\Rightarrow$ Sensing & Perception.
• Where am I going? $\Rightarrow$ Localisation & Mapping.
• How should I get there? $\Rightarrow$ Planning & Control.

Assessment

• Three-hour exam (80%), answer 4/7 questions, same format as past papers.
• Coursework (20%), detailed below.

Coursework

A: Paired practical activity over 5 weeks [20%]
B: Individual written assignment (approx. 2000 words) [80%]

You will be programming a Lego NXT robot using the LeJOS firmware (writing your solutions in Java).

Deadline: easter holiday.

Programming Prerequisites

• Basic algebra: vectors, matrices, …
• Some extra calculus: vector calculus, differential equations, …
• Probability theory: probability distributions, Bayes rule, …
• Programming: Java.

Suggested Reading

• Sebastian Thrun, Wolfram Burgard, Dieter Fox, Probabilistic Robotics, MIT Press

What is a Robot?

A robot is:

1. An artificial device that can sense its environment and act purposefully in it.
2. An embodied artificial intelligence.
3. A machine that can autonomously carry out useful work.

Building on that we have mobile robots:

• A mobile robot is a robot capable of moving around in its environment and perform certain tasks.
• The ability to move extends the workspace of a mobile robot substantially, an attractive feature for many applications.
• However, extended workspace also means that the mobile robot should be able to do the following: have some knowledge about its environment, recognize objects in the environment, generate a real-time response, and do all of these simultaneously.