You may be interested in the work I’ve recently done on extending chunks & rules to support swarms of cognitive agents. Agents collaborate through message exchange and mechanisms for task synchronisation. Each agent may be responsible for multiple devices via their digital twins.
Agent to agent chunk messaging: either to named agents or via topic-based streams. This is decoupled from the underlying protocols which need to support reliable, timely in-sequence delivery, e.g. Zenoh, MQTT, DDS, WebRTC, HTTP and Web Sockets.
Tasks as an abstraction for named threads of behaviour: this enables task synchronisation, e.g. where one task needs to be done only when several others have completed. Agents can also delegate tasks to other agents. Tasks also provide a basis for reinforcement learning, although I have yet to implement that.
You can find slides from my lecture to the Edge AI summer school in Pisa, a paper I submitted to GIECS 2025 and a couple of demos. I next need to work on extending the chunks & rules test suite and updates to the spec.
Comments welcomed!
https://github.com/w3c/cogai/tree/master/demos/Swarms/tasks
cogai/demos/Swarms/tasks at master · w3c/cogai
github.com
Dave Raggett <dsr@w3.org>
