Draft document on dynamic scheduling - Tony Wong, ATNF 12 Nov 2001 ====================================================== Dynamic scheduling adjusts the observing schedule in real time according to system status and observing conditions. For example: 1) Point source projects are favoured when antennas are missing 2) Millimetre projects are favoured when weather is good Currently the Compact Array is not well suited for dynamic scheduling, since many CAOBS commands cannot be executed within SCHED files. There are also observing-related tasks like CACAL, CATAG and SPD which must be run outside of CAOBS. Thus, a substantial software effort will be required to allow automated observing. The advantages of dynamic scheduling are clear: the most efficient possible use of the array. Indeed, the experience of the past year suggests that without dynamic scheduling, very little useful millimetre science will be possible at Narrabri. ------------------------------------------------------ A PROPOSAL FOR IMPLEMENTING DYNAMIC SCHEDULING AT ATCA ------------------------------------------------------ Observing: ---------- 1. Proposers will still be responsible for creating schedule files. They may be present at Narrabri if they wish, and will certainly be encouraged to come if they are unfamiliar with the instrument, but they will have no control over time allocation. 2. The DA is responsible for making sure the scheduling software is working. He/she is not responsible for ensuring that schedule files are properly written, although help should be provided when practical. If a schedule file has errors and cannot be observed, and the proposer is not present at Narrabri or monitoring observations remotely, then another project will be observed instead. 3. Because the responsibilities of the DA are increased, it may be sensible to have 2 DA's at any given time, with alternate shifts. 4. Certain (technically difficult) projects may be designated as "interactive only", i.e. they may still be dynamically scheduled, but the proposer must observe the project either at Narrabri or remotely. For dates on which the PI promises in advance to be "on call", and on which observing conditions are appropriate, the software simply relinquishes control of the array to the observer. On other dates, the project will not be observed. 5. Certain (high-priority cm) projects may be designated "fixed allocation", i.e. the observing times are fixed and cannot be overridden. Obviously the number of such projects should be minimised for dynamic scheduling to be effective. 6. Traditional, fixed-queue scheduling may remain for array configurations in which there are no high-frequency proposals. However, to reduce confusion, these observations would still be handled by the scheduling software, unless designated "interactive." Software Requirements: ---------------------- 1. Automatic source selection: User should be able to specify LST ranges for sources and prioritize calibrators. In other words, they should be able to create a schedule that is robust against changes to starting times/dates. 2. Hands-free operation: CACAL should not need to be run interactively. Flux or bandpass calibrators should be included in the same schedule file as the source and phase calibrator. Decision making: ---------------- Project requirements will depend on: * Frequency * Desired system temperature * Desired rms phase (e.g., using phase referencing or selfcal) The decision on whether to stop observing current project will depend on: * Project requirements * How often the observing conditions are re-assessed * Minimum time a project must be allowed to run. The decision on which project to observe next will depend on: * Project requirements * Priority * Number of allocated hours completed * HA ranges completed and needed Social impact: -------------- Implementation of dynamic scheduling will reduce the number of visitors to Narrabri. This can be compensated to some extent by encouraging first-time PIs to be present at the array and doubling the number of DA's. We might also increase the quality of staff-scientist (and scientist-scientist) interactions by having an annual science festival on-site. Dynamic scheduling may also reduce the number of visitors to Australia. However, the greatest impact will likely be on repeated visits by overseas Australians or others familiar with the ATCA. Other foreigners would probably want hands-on experience with the instrument, or would ask a local collaborator do the observing anyway. Ultimately, the increased productivity of ATCA should lead to better science and hence more visitors. Dynamic scheduling may increase the number of overseas proposals from those who cannot afford to visit Australia for observations. This would increase the visibility of ATCA, foster more collaborations with Australian astronomers, and raise the standard of ATCA projects. Implementation: --------------- Phase One (2002) will involve continuously monitoring observing conditions and running a dynamic scheduling program in simulation mode. An ATNF-led committee will meanwhile develop software for dealing efficiently with dynamically scheduled observations. We will then assess the results of the simulation in mid-2002. In Phase Two (starting in mid to late 2002), the array will actually be scheduled dynamically, at least in certain array configurations. In practice, we might expect the balance between fixed, interactive observing and dynamic scheduling will shift gradually, as users become more accustomed to the new system.