| Segregating Planners and Their Environments (2007) | |||||||||||||
Abstract | |||||||||||||
| By implementing agents and environments using a domain-independent, extensible simulation substrate, described in this paper, agents will have clean interfaces to their environments. These makes it easier for agents to be plugged into other environments that have been similarly defined. If agents can interact with multiple environments, their behaviors and the associated experimental results will be more general and interesting. 1 Introduction We argue against a tight integration of planners and their environments. This is not because we think integration is a bad thing, but because we think it's important to run planners in a variety of environments. An AI Magazine article [6] describes a number of AI environment simulators, including costs and benefits. By running an agent in a variety of environments, we can hope to get the benefits of all worlds. To make it easier to mix and match planners and environments, they must have compatible interfaces. The worst case is illustrated in fig... | |||||||||||||
Publication details | |||||||||||||
| |||||||||||||