VOEventNet: Real-Time Astronomy with a Rapid-Response Telescope Grid

Project Description

Exploration of transient sources represents one of the last frontiers in modern astrophysics. The development of a comprehensive understanding of a new event requires real-time observation with multiple instruments. Yet while telescope facilities continue to ramp up to unprecedented data rates, there has been no concerted effort to ensure real-time communications of astronomical events. A federated response must be enabled to push transient astrophysics ahead in the 21st century.

VOEventNet is a peer-to-peer cyberinfrastructure to enable rapid and federated observations of the dynamic night sky. VOEventNet is a network of telescopes and computers working synergistically, under the watchful eye of humans, to find and study interesting astronomical events. It includes a common language for describing transient events, generation of event streams from current surveys, publication, archiving and persistent identifiers for events; and transportation of events to interested subscribers, with all of this happening automatically in seconds or minutes after discovery.

The architecture of VOEventNet. The Palomar-Quest survey runs a real-time event factory on TeraGrid in streaming mode (red arrows, left), which finds changes in the sky by comparison with previous observations. The Event Synthesis Engine uses all observational data from PQ and robotic telescopes, plus remote database services from the Virtual Observatory (NVO). Events are published to VOEventNet, with federated repositories (green) at Caltech and Los Alamos . Immediate, automated observations from robotic telescopes feed back into the network, allowing an event to be cycled back into the synthesis engine again.

This activity is a synergy of distributed computing with distributed sensors. Though set up by humans with pre-defined criteria, VOEventNet operates automatically and autonomously. We will build an event factory to detect sky transients in near real time, and utilize novel machine-learning technology to distinguish scientifically relevant transients from false positives through synthesizing multiple observations. In this way, VOEventNet will enable an already active area of astrophysical research to flourish. The system is not just research with the hope of a prototype; this is the unification of existing powerful resources to form a new, robust, competent production system that will generate a great deal of new science.

VOEventNet will be a driving force for a rapidly growing new area of astronomical inquiry, eventually serving as a transparent backbone to enable pan-facility communication. A broader output of this project concerns the distributed decision making based on synthesizing heterogeneous, multi-sourced data, including publishing observations of events that includes an importance/urgency rating for follow-up, so that subscribers can decide independently whether to take action, based on their own criteria. Another crucial question is how trust can be engendered and maintained in a sensor-computer network that is pre-programmed with criteria, and follows that program quickly and unexpectedly. As a complex real-world distributed network, it serves as a case study in the development of a dynamically data-driven application, which has the potential to bring a transformative change to science, and open entirely new domains of parameter space for scientific investigations. The project will involve training of students and postdocs and broader public outreach, contributing to the education of the next generation of science and technology leaders.