Current Software

Intelligent Agents

A simple block diagram showing how the user would, in the simplest case, make a direct observation request. Here an observation request is made by the user, and distributed by the user's agent to all agents embedded at telescopes on the network, who then score the request and return their evaluation of how well they could perform the observation to the user's agent. The user's agent then chooses the telescope best able to carry out the observation and places an observation request with than telescope. Once accepted, the observations are queued and carried out by the telescope with the data products being returned by the embedded agent to the user's agent. The agent may either then return these to the user directly, or possibly carry out autonomous follow-up depending on the results of the observation run.

The current generation of eSTAR software is based on a SOAP transport layer, with a far more modular approach being taken to the architecture.

The paradigm used by the prototype was that of the multi-agent contract model. In such a architecture both the software controlling the science programme, and that controlling the telescopes, are thought of as agents. With a contract being negotiated between the two for services to be provided. However, as the project progressed it become clear that the multi-agent contract model is largly inappropriate.

The paradigm used by the current generation of intelligent agent software is that of the collaborative agent model, where several agents pool their expertise to solve a problem. What makes the collaborative agent model so attractive is that we can split the science agent itself into several smaller agents, each of which can provide web or grid services.

This increases the code reusability and flexibility of the system with very little overhead, and allows the user to run different top end interfaces into the collaborative agents.

Both command line and graphical user interfaces into the user's science agent are available, and the project is looking at the possibility of having a Java based interface which could run on the user's mobile phone. This would provide up to the minute information for an astronomer who is out of the office, but still wants to monitor their observations.

The software has been deployed onto the United Kingdom Infrared Telescope (UKIRT) on Mauna Kea and onto the Robonet-1.0 network of telescopes and has been operating for over a year performing rapid followup of Gamma-Ray Bursts on UKIRT, and hunting for exo-planets using Robonet-1.0. The current network status of the deployed eSTAR software can be monitored live online.