DevelopmentalRobotics - Emergent Wiki

1. Developmental Robotics

Developmental Robotics is the name of a research program started by Lisa Meeden, Deepak Kumar, and Douglas Blank.

Blank, D.S., Kumar, D., Meeden, L., and Marshall, J. (2004). Bringing up robot: Fundamental mechanisms for creating a self-motivated, self-organizing architecture. Expanded and revised version of a conference paper, under review for a special issue of Cybernetics and Systems.
Stober, J., Meeden, L., Blank, D.S. (2004). The Governor Architecture: Avoiding Catastrophic Forgetting in Robot Learning, under review.
Blank, D.S., Kumar, D. and Meeden, L. (2002). A Developmental Approach to Intelligence. In Proceedings of the Thirteenth Annual Midwest Artificial Intelligence and Cognitive Science Society Conference, Edited by Sumali J. Conlon.
Blank, D.S., Kumar, D. and Meeden, L. (2002). Bringing up robot: Fundamental mechanisms for creating a self-motivated, self-organizing architecture. In Proceedings of the workshop Growing Up Artifacts that live, Simulated Adaptive Behavior 2002, From Animals to Animats.
Blank, D.S., Kumar, D. and Meeden, L. (2002). A Developmental Approach to Anchoring. Bryn Mawr College Computer Science Technical Report 2002-01.

Pyro is the project being developed to support research into Developmental Robotics. You'll find everything you need at the Pyro web page.

The following sections roughly outline several activities for this summer.

A log of specific activities is kept at DevelopmentalRoboticsSummerLog.

Please record all your activities on the page above.

A log of Evan's thesis research and Jeremy's directed reading/research can be found at GovernorResearchProgramLog.

SOM_PAK: The Self-Organizing Map Program Package Documentation
Self-Organization and Autonomous Robots, J. Heikkonen and P. Koikkalainen
Self-Organizing Maps, Third Edition, T. Kohnen, Chapter 3
Introduction to the Theory of Neural Computation, J. Hertz, A. Krogh, and R.G. Palmer, Chapter 9: Unsupervised Competitive Learning,

Proceedings of Workshop on Development and Learning
Autonomous Mental Development by Robots and Animals, J. Weng, J. Mc Clelland, A. Pentland, O. Sporns, I. Stockman, M. Sur, and E. Thelen
Toward Bootstrap Learning for Place Recognition, B. Kuipers and P. Beeson
Developmental Robotics: Theory, Method and Experimental Results, J. Weng, W.S. Hwang, Y. Zhang, and C.H. Evans
SAIL and Dav Developmental Research Projects
Bringing Up Robot: Fundamental Mechanisms for Creating a Self-Motivated, Self-Organizing, Architecture, D. Blank, D. Kumar, and L. Meeden
Made-Up Minds, G. Drescher
Integrating emotion and motivation into intelligent systems, N. Alvarado, S. Adams, S. Burbeck, and C. Latta
The Symbol Grounding Problem, S. Harnad

Compare four methods of creating neural network representations of SOM results: 1) winner x,y 2) winner list 3) gaussian list 4) error list

Run 5 sets of experiments for each representation at both sites
SOM parameters: initialize randomly, use default decay method, train for 5000 steps
SOM sizes: motor 4x3, ir and light 8x6, vision 12x8
Network parameters: lrate 0.5, momentum 0.1, hidden layer size is set to half the size of the input layer
Use a rectangular environment with a long barrier in the center and a light in one corner
Train until the network is predicting the next SOM winners correctly 75% of the time

Re-organize the wiki wiki documentation of pyro. See PyroDoc
Document the data collection and visualization processes
Low-level, threaded vision drivers in python
Fix the queuing mechanism in python code for the serial line of the khepera to separate the writing and reading aspects
Add a thread to handle openGL redraws in pyro