Course content

Topics in this course include: bio-inspired algorithms (evolutionary algorithm, artificial neural-network, swarm algorithm, etc), bio-inspired sensory-motor control (central pattern generator, behavior-based architecture, etc), morphological computation, etc.

Each topic will be covered in lectures (theoretical concepts and research case studies) and exercises (3 during class and 1 take-home). We will use quadrupedal (four-legged) robots as a platform to gain hands-on experiences and skills to implement various biologically inspired algorithms and design approaches. Each student group (of 3-5 people depending on class size) will have one robot to work on for the class project (see more info below).

Note: The exercises will include programming tasks.

Class Project

Students will work in groups to apply the biologically inspired concepts presented in lectures to improve their robot's locomotion performance, i.e. to be able to walk/run faster, display a larger range of walking gaits, etc. The exercises will provide all of the required tools to implement various algorithms and approaches.

For the project, each student group can choose their favorite algorithm and design approaches to implement on their robot. At the end of the course, all student groups will participate in a competition to evaluate their robot's locomotion abilities. This competition will include activities such as: robot speed racing, creative walk/run/dance styles, etc.

Each student group must also give a short presentation and a written report.

Grading

There will be no final examination. Instead, the course grade will be based on the class project (short presentation and written report).

In addition, each student must complete at least 50% of the 4 exercises in order to participate in the class project and receive a grade.

Tentative Schedule

17 September: Lecture 1: Introduction (slides) and arduino_tutorial

24 September: Lecture 2: Bio-inspired robotics: locomotion (slides) and webots_tutorial

1 October: In-class exercise 1

8 October: Lecture 3: Bio-inspired robotics: morphology and sensory-motor control (slides) and Programming_the_robot I

15 October: Lecture 4: Evolutionary algorithms I (slides) Programming the robot II

22 October: Lecture 5: Evolutionary algorithms II & Neural network algorithms I (slides) Programming the robot III

29 October: In-class exercise 2

5 November: In-class exercise 3

12 November: Lecture 6: Neural network algorithms II (slides)

19 November: Lecture 7: Neural network algorithms III (slides)

26 November: Lecture 8: Swarm algoriths, everything else and wrap up (slides)

3 December: Project preparation

10 December: Robot competition

17 December: Project presentation

Lijin Aryananda, email = lijin at ifi.uzh.ch

Literature

Dario Floreano, Claudio Mattiussi, "Bio-inspired Artificial Intelligence: Theories, Methods, and Technologies", MIT Press, Cambridge MA, US, Sept 2008. ISBN: 978-0-262-06271-8.

Reading materials

For lecture 2 and 3:

1. R. Pfeifer et al, Self-organization, embodiment, and biologically inspired robotics, Science 318, 1008, 2007 link (you may have to be on the university network to access this)

2. R. Full, Using biological inspiration to build artificial life that locomotes, Proceedings of the International Symposium on Evolutionary Robotics From Intelligent Robotics to Artificial Life, 2001 link

3. F. Delcomyn, Insect walking and robotics, Annual Review of Entomology, Vol. 49: 51-70, 2004 link (you may have to be on the university network to access this)

For lecture 4:

Textbook chapter 1

For lecture 5:

Textbook chapter 3

Competition rules (to be posted soon)

1. Evolutionary algorithm for finding optimal gaits and body parameters.

Research examples:

G. Hornby et al. Autonomous Evolution of Gaits with the Sony Quadruped Robot. Proceedings of 1999 Genetic and Evolutionary Computation Conference, 1999. (pdf)

G. Hornby et al. Autonomous Evolution of Dynamics Gaits with Two Quadruped Robots. IEEE Trans on Robotics 21:3, 402-410, 2005. (pdf)

2. Subsumption architecture

R. Brooks. A Robot that Walks; Emergent Behaviors from a Carefully Evolved Network, Technical Report, MIT 1989. (pdf)

3. Insect inspired reflex-based control

H. Cruse et al. Walknet – a biologically inspired network to control six-legged walking. Neural Networks 11: 1435-1447, 1998. (link)

Application of this method on a bipedal robot:

(shorter version) P. Manoonpong et al. The RunBot Architecture for Adaptive, Fast, Dynamic Walking. ISCAS, 2007. (pdf)

(longer version) T. Geng et al. Fast Biped Walking with a Sensor-driven Neuronal Controller and Real-time Online Learning. International Journal of Robotics Research vol 25, no 3, 243-259, 2006. (pdf)

If you are interested in implementing any of these approaches on your robot, please let us know and we will help you with the implementation. We are currently working on providing you with an implementation of evolutionary algorithm, combined with Webots.