Previous abstract Next abstract
This abstract describes our project on scheduling for networks of remote robotic telescopes. The project is being developed as part of our larger goal to build automated tools to address the complete life-cycle of an observation request, from electronic transmission of the observation request to the return of raw and reduced data, using the Automatic Telescope Instruction Set, or ATIS. With distributed artificial intelligence (DAI) software techniques, we have designed a network scheduling system as a collection of distributed, cooperating scheduling agents. Each agent is responsible for the scheduling of observation requests for one robotic telescope. This perspective allows the network scheduling system to preserve the individualized priorities and policies that may exist for any one telescope while promoting the collaborative behavior required for acquiring and providing telescope time in a network. We are interested in heterogeneous networks that are made by connecting pre-existing fully automated remote telescopes. Providers of telescopes are anyone with a fully automated telescope and telecommunication capabilities. In essence, the network system is created by interconnecting multiple stand-alone systems. The scheduling software for each stand-alone system (or agent) is based on two primary components: an advanced scheduling system, CERES, that employs look-ahead contingent scheduling methods, and a collaborator, that communicates with other agents in the network, both written in C and Lisp. Cooperation among agents is based on the premise that astronomers at one telescope are often willing to trade for time on another telescope. In general, agents are autonomous but may react cooperatively to observational requests communicated by other scheduling agents. Our work involves engineering the protocols for cooperation, and the development of a programming language and agent architecture to express such protocols. The system is being constructed with the help of a Multi-Agent Test Environment used for monitoring and coordinating distributed multi-agent systems on local area networks. Agents operate on Sparc (UNIX) workstations, with telescope controllers on PCs.
Thursday program listing