Intelligent Robotics: Introduction

Transcription

1 Intelligent Robotics: Introduction Intelligent Robotics Intelligent Robotics (Extended) Jeremy Wyatt School of Computer Science University of Birmingham, 2011/12

2 Plan Intellectual aims of the module Task description Introduction to hardware and software Using robots to understand intelligence

3 Module aims 1. Give an appreciation of the issues that arise when designing complete, physically embodied autonomous agents. 2. Introduce some of the most popular methods for controlling autonomous mobile robots. 3. Give hands on experience of engineering design. 4. Encourage independent thought on possible cognitive architectures for autonomous agents.

4 Learning outcomes 1. Design and program simple autonomous robots. 2. Implement standard signal processing and control algorithms. 3. Describe and analyse robot processes using appropriate methods. 4. Write a detailed report on a robot project. 5. Carry out and write up investigations using appropriate experimental methods.

5 Basic Hardware DC motors Servo motors Odometers Sonar

6 Hardware: a warning 1. Please take extreme care in handling of all hardware 2. If you are not 100% sure then ASK 3. You may not remove laptops from the lab. They are part of the robot. 4. Robots may operate in the lower ground floor, but be careful with drinks, people etc. 5. Total kit value = 50k. This a major investment solely for your benefit.

7 Assessment Exercise 1: Getting started (5%) Exercise 2: Let's go for a walk (5%) Exercise 3: Where am I? (10%) Exercise 4: Probabilistic Road Maps (20%) Exercise 5: Call a meeting (60%) /exercises/

8 Assessment Dates Exercise 1: 4th October (viva in lab) (5%) Exercise 2: 11th October (viva in lab) (5%) Exercise 3: 25th October (viva in lab) (10%) Exercise 4: 8th November (viva in lab) (10%) Exercise 5: Demonstration on 6th December, (20%) Exercise 5: Report hand-in to School Reception at 12 noon on 10th January 2012 (50%)

9 What s Easy is Hard Easy: expert systems, mathematics, chess Hard: seeing, language understanding, moving around, making a cup of tea, common sense What s easy for humans is hard for computers and vice versa. Why?

10

11 The Whole Iguana AI commonly studies aspects of intelligence separately: narrow domain high performance In 1976, philosopher Dan Dennett suggested putting it all together, but with a low level of performance

12 The Whole Iguana... why not obtain one's simplicity and scaling down by attempting to model a whole cognitive creature of much less sophistication than a human being?... a turtle, perhaps... [but] considering the abstractness of the problems properly addressed in AI... one does not want to bogged down in the cognitive eccentricities of turtles if the point of the exercise is to uncover very general, very abstract properties that will apply as well to the cognitive organisation of human beings. So why not make up a whole cognitive creature, a Martian three-wheeled iguana, say, and an environmental niche for it to cope with? I think such a project could teach us a great deal about the deep principles of human cognitive psychology

13 Experiments with vehicles Behaviour of agents was more complex than their mechanisms Behaviour depended on the environment as well as the agent Hard to infer mechanism from behaviour alone

14 Experiments with vehicles Valentino Braitenberg - Law of uphill analysis and downhill invention My Conclusion: synthesizing agents may have something to offer in understanding our minds

15 Why build robots to understand minds? All naturally occuring intelligence is embodied So robots are in some ways similar systems Robots, like animals exploit their environments to solve specific tasks There are no general purpose animals why should there be general purpose robots? David MacFarland

16 Lessons from nature Gannets wings half open to control dive, and fold wings to avoid damage Fold not at a constant distance, but at a constant time to impact Birds have detectors that calculate time to impact

17 Task specific robots Polly the tour guide exploits assumptions about environment to perform task quickly

18 Lessons from nature: 2 Other animals are capable of a surprising degree of manipulative ability e.g. Betty the crow who can make tools Sometimes we can use robots to test theories of how specific animals work e.g. cricket phonotaxis

19

20

21

22 Wrap up By synthesising intelligent robots we can address deep questions about the nature of intelligence Robots, like animals, are embodied We can use the task-environment dynamics to constrain our computational problems