[Modeling] about modeling environment in agent-based simulation

James Odell email@jamesodell.com
Sun, 29 Feb 2004 13:32:52 -0500


Are there any particular papers at your website that you could recommend?


On 2/29/04 7:53 AM, Danny Weyns scribed:

> Hello Ye,
> your question concerns an important issue in agent-based systems. In my
> opinion the environment need to be treated as a primary abstraction in
> multi-agent systems, without any doubt!
> In most models for mas, the environment is reduced to communication
> infrastructure. However, this neglect the true potential of the
> environment. Here are a couple of questions of my favourite list to
> illustrate the necessity of the environment as an abstraction on its own
> right:
> -how to deal with objects or data in the environment
> -what about the actions (possibly simultaneous actions) of agents onto
> these objects
> -how to control the effects of the agents' interactions independently of
> the internal architecture of the agents
> -how to deal with the (dynamic) topology of environments
> -what about processes in the environment that do not directly
> relate to agent activity such as the variation of an environmental
> variable (e.g. temperature), the expansion of a field in the
> environment, the movement of an object or the evaporation of a
> pheromone?
> We need the environment as an abstraction to model these issues, it seems
> unnatural to me to put it all in the agents?
> In my work, environment is treated as a primary abstraction, please find
> my papers at
> http://www.cs.kuleuven.ac.be/~danny/publications.html
> Note that at the next AAMAS conference a workshop is dedicated to
> environment for mult-agent systems, see
> http://www.cs.kuleuven.ac.be/~distrinet/events/e4mas/
> with regards,
> Danny Weyns