Artificial Models of Natural Intelligence

Bath AI  |  Intelligent Systems Group  |  Contact Us 


We try to put all code online, particularly for papers that are already published. Please feel free to contact individual researchers if you can't find enough information below. Unless stated otherwise, all code below is open source under the terms of the MIT License. The copyright is held by the code's author(s). For more information, see Joanna Bryson's publications pages, and the AmonI Research page

Behavior Oriented Design (BOD) and POSH Action Selection

Behaviour oriented design is a methodology for developing intelligent systems.  It extends object oriented design to the special problems of proactive systems, including real-time systems for dynamic enviornments.  To the extent that these systems are agents, they need goals and priorities; in BOD these are specified using POSH action selection.

  • Behavior Oriented Design home page. (No code, just explanations.)
  • POSH dynamic planning home page (explanations & specification).
  • There is an IDE for BOD, A BOD Environment (ABODE)
    • The present version is ABOD3 provides real-time debugging as well as editing. It works across many different representations of POSH, including Instinct (see below.) by Andreas Theodorou
    • ABOD3-AR is an extension of ABOD3's real-time debugging mode that works as augmented reality on your smartphone for debugging mobile robots on the run. code by Alexandros Rotsidis
    • Older versions are just editors:
  • Downloadable engines for running POSH:
    • A game-centric version of POSH has been used in the development of BUNG (BOD UNity Game); we have been using BUNG in our final-year and MSc-level course, Intelligent Control and Cognitive Systems.
    • Instinct is a reactive planner based on POSH. It includes a visual design language for plan design, and enhancements to support transparency.
      • It is written in C++ and has been designed to operate within a tiny memory footprint such as that available on the R5 Robot.
      • It has been tested with the Arduino AVR microcontroller and also Microsoft Visual C++.
      • Instinct integrates with the ABOD3 IDE.
  • Please feel free to become a BOD Developer
  • Older projects that are no longer maintained but may still be useful:

Models of Social Learning and Cultural Evolution

Realistic Emotion Representation

  • Emmanuel Tanguy created the Dynamic Emotional Representation (DER), a natural, real-time models of emotions for facial animation. This is distributed as a part of the Emotionally Expressive Facial Animation System (EE-FAS). updated July 2006
  • Philipp Rohlfshagen working with Joanna Bryson created a model of emotions and drives that is relatively realistic but also tractable for ordinary action selection. This is distributed with standard pyPOSH version of BOD code, above. See the latchTest library for examples and documentation.

Individual Learning

  • Mark Wood's PhD was on "An Agent-Independent Task Learning Framework". dissertation and code in Java, including a framework for running social learning experiments in Unreal Tournament (UT).
  • Joanna Bryson's model of task / transitive inference learning:
    • Lisp code that generated the results in Animal Cognition (Bryson & Leong 2007).  This is its own simulation, that page points to the lisp environment the GUI uses.

Models of Social Behaviour