Course Outline - Semester 1, 2008
For more details see the Official Course Profile.
Presenters:
Course Coordinator: Dr Gordon Wyeth
Phone: 3365 3770 Email: g.wyeth@uq.edu.au
Building: Axon Building Room: 508
Lecturer: Professor Srini Srinivasan
Phone: 3346 6322 Email: m.srinivasan@uq.edu.au
Building: Queensland Brain Institute Room: 335
Lecturer: Professor Janet Wiles
Phone: 33652902 Email: j.wiles@uq.edu.au
Building: Axon Building Room: 506
Objectives:
Studies in how to design robots based on ideas from biology, and to make robots that behave like biological organisms. Areas of study include movement, evolution, learning, vision and navigation. Practical work using simulations.
By successfully completing the course students should be able to:
- Explain the state of the art in a range of biologically inspired robotic systems
- Design simple robot systems based on ideas from biology
- Implement simple models of biological systems using robots and simulations of robots
- Appreciate the benefits and limitations of looking at biological systems for inspiration in engineering design
- Appreciate the benefits and limitations of building engineering models to better understand biological systems
- Understand the relationships between the distinct vocabularies and research approaches of biological and technical researchers
Assessment:
Braitenberg Vehicle Simulation (20%):
Use a Braitenberg Vehicle Simulation system to make and study an ecosystem of
animats. Assessment is based on a 2000 word report on the simulation study.
Due Date: 16 Apr 08 16:00 - 16 Apr 08 16:50 Late assignments will lose 5 marks.
They may be handed in at the following tutorial.
Organism Case Study (30%):
Choose a biological organism and rigourously investigate how it has been used or
might be used as the basis of a robotic system. Assessment is based on a 3000
word report on your research.
Due Date: 21 May 08 16:00 - 21 May 08 16:50 Late assignments will lose 7.5
marks. They may be handed in at the following tutorial.
Annotated Bibliography (50%):
Put yourself in the position of a reviewer for each paper in the Reading List.
Write a review for each paper (250-500 words for each paper).
Due Date: 10 Jun 08 09:00 - 10 Jun 08 17:00
Course Outline:
Week 1: Introduction (Gordon Wyeth)
- Course outline and policies
- Bio-inspired engineering design
- Engineering models for biology
- Crossing the Bio-Techno Divide
- Ethics of Biological research
- The Importance of a Body
- Taxis and Kinesis
- Case studies:
- Braitenberg Vehicle 1
- E-Coli
Week 2: Sensori-Motor Integration (Gordon Wyeth)
- Neuron Models
- Neurons in Bodies
- The Two Neuron Trick
- Synthetic Psychology
- Swarms and Communities
- Case Studies:
- Braitenberg Vehicles 2-3
- Nematode Worm
- Crickets
Week 3: Walking and Swimming (Gordon Wyeth)
- Principles of Walking
- ZMP Control
- Central Pattern Generators
- Natural and Artificial Actuators
- Case studies:
- Cockroaches
- Lamprey / Salamander
- Quadruped Robots
- Biped Robots
Week 4: Learning (Gordon Wyeth)
- Plasticity in Neural Systems
- Supervised Learning
- Concept Learning
- Motor Learning
- Operant Conditioning
- Case studies:
- Aplysia
- Braitenberg Vehicle 7
Week 5 : Natural Evolution (Janet Wiles)
This week we will briefly look at biological life and the evolution of life on earth. We will review natural evolution, and the contributions made by variation and selection. Case study: evolution of developing organisms.
Week 6 : Artificial Evolution (Janet Wiles)
This week we turn to evolutionary computation, covering artificial genomes, mutation operators and fitness functions. Evolution is a subtle process, for example, in the ways that learning and selection interact, a process known as the Baldwin effect and in interactions between variation and development. Case study: evolutionary robotics.
Weeks 7 and 8: Navigation (Gordon Wyeth)
- SLAM: Simultaneuous Localisation and Mapping
- Probabilisitic Solutions to SLAM
- Biological Studies of SLAM
- Attractor Models of Spatial Encoding
- Cognitive Maps
- Case Studies
- Rodents
- RatSLAM Robots
Week 9: Insect Vision: Stabilisation and steering mechanisms (M.V. Srinivasan)
Here we take a look at the ‘cockpit’ of a fly to unravel the processes by which an insect maintains stable flight and a constant heading (if it so chooses), as well as the mechanisms that enable the visual detection and pursuit of potential mates.
Week 10: Insect Vision: 3-D vision (M.V. Srinivasan)
Most insects have poor or no stereo vision. Nevertheless, they seem to have evolved alternative strategies, based on image-motion cues, to estimate the distances to objects and surfaces, avoid collisions with them, and perceive the world in three dimensions. Here we examine how this is done.
Week 11: Insect Vision: Navigation (M.V. Srinivasan)
Ants and bees are impressive navigators. They use the sun and the pattern of polarised light that it creates in the sky as a ‘celestial’ compass, and navigate by combining path integration (dead-reckoning) with landmark information. Here we review these strategies and the underlying mechanisms.
Week 12a: Insect Vision: Perception and ‘cognition’ (M.V. Srinivasan)
Bees learn and recognise not only the colours of flowers, but also their shapes. While there is much evidence that patterns are memorised in an eidetic (‘photographic’) way, recent research suggests that bees can also abstract general features of patterns and apply the information to distinguish between patterns that they have never encountered before. Bees use ‘top-down’ processing to detect camouflaged objects, and can learn to negotiate complex mazes. Some of these faculties are reminiscent of those possessed by higher vertebrates.
Week 12b: Insect-inspired robots and seeing machines (M.V. Srinivasan)
Traditionally, machine vision has been approached from "first principles" (i.e. mathematics), with relatively little consideration as to how the problems that are tackled might be solved by natural visual systems. While the "first principles" approach is a perfectly valid one, the solutions that it produces often tend to be computationally complex, sensitive to noise, and sometimes too general-purpose to be really effective. On the other hand, it is patently clear that animal vision is rapid, reliable and robust. Here explore alternative strategies for artificial vision that take advantage of some of the computational "short-cuts" that insects seem to have evolved.
Week 13: Panel Discussion and Review (All lecturers)