Long-Lived Agents on the Web: Continuous Acquisition of Behaviors in Hypermedia Environments
Danai Vachtsevanou, Philip Junker, Andrei Ciortea, Iori Mizutani, Simon Mayer · Companion Proceedings of the Web Conference 2020 · 2020
We demonstrate how autonomous goal-directed agents can exploit hypermedia to acquire and execute new behaviors at run time. In addition to behaviors programmed into the agents, in our system agents can discover and reuse behaviors extracted from machine-readable resource manuals. Such manuals can be published by developers, synthesized by agents through automated planning, or even specified by human users at run time. Agents can then discover and use physical and virtual resources in flexible ways, which allows them to better cope with the rapid evolution of open and dynamic Web environments.