1. OpSim4 vs OpSim3 Francisco Delgado 2017-07-05

2. OpSim evolution OpSim1 Proof of concept in IDL
Simulation of visits with multiple science cases.
OpSim2
Python
Detailed model for the observatory and the weather
Embedded Scheduler prototype
Telescope design validation, site selection
OpSim3
Additional science cases and scheduling algorithms
Modular Scheduler prototype
Parameters exploration, survey definition validation

3. OpSim3 vs OpSim4 Single process SOCS process + Scheduler process
DDS interface between SOCS and Scheduler
Direct python calls between telemetry simulation and scheduler
Single observatory model
2 instances of observatory models
Only SOCS writes DB
Multiple DB access
Configuration messages
Configuration files

4. OCS communications Middleware
Scheduler in Control Software
OCS Operator
OCS Remote
OCS Monitor
OCS Application
Control
Scheduler
History
OCS
EFD
OCS Sequencer
Telemetry
Image Quality
Cmd
Visits
Targets
Sched
Telem
Visits
OCS communications Middleware
TCS
CCS DM

5. OCS communications Middleware
Scheduler in OpSim4
Control
Scheduler
History
Telemetry
Image Quality
Targets
Sched
Telem
Visits
OCS communications Middleware
SOCS

6. Scheduler Internal Block Diagram
Time
Main
Driver
Sched Config
Scheduler
Sched Mode
Control
Targets
Targets
Downtime
Cost functions
Sched Telem
Degraded
Sched Telem
Slew Time
Observatory Model
Telemetry
Observatory conditions
Sky
Model
Pre-Calc Data
Candidates
Weather forecast
Proposals
Image Quality
Environment conditions
Value functions
History
Quality parameters
Observation History
Past observations
Visits
Current observation 6

7. Accuracy changes Repeatability
OpSim4 offers 100% repeatability in observations sequence and slew states. OpSim3 was not.
Alt-Az precision improved
Alt-Az is estimated at slewInit time and is recalculated at slewFinal time.
Difference can be seen between “target” and “slewFinalState”. OpSim3 was not recalculating.

8. Accuracy changes Remaining tracking time in OpSim4
Taken into account before sending target (short term look-ahead).
Avoid reaching tracking limits in altitude, azimuth or rotator.
Ignored in OpSim3
Sky brightness model
New model, per band, includes explicit twilight
No caching ranks (reuse)
Everything is ranked at each visit

9. Area Distribution in OpSim3

10. Area Distribution in OpSim4
Simpler integrated ranking in OpSim4 promotes a smoother distribution in all filters for the specified area in comparison to the 2 steps ranking in OpSim3

11. Area Distribution changes
New Airmass bonus
New Hour Angle bonus

12. Time Distribution changes
OpSim3: parameters use reference midTime=0
OpSim4: parameters use reference midTime=desired interval

13. Area-Time hybrids OpSim3 hybrid is in Modified Time
The first of a group is ranked in Area only (WLprop)
OpSim4 hybrid is in Modified Area
Optional configurable grouped timed visits
The first visit in the group is area ranked only
The following visits in the group are area ranked and time-window modulated

14. Area Distribution changes
Same night revisits constraint
Revisits to the group during the same night can be avoided. For example, no more than 2 visits per target per night.
OpSim3 did not have the parameter to constrain revisits in the same night.
No overflow
Smoother distributions in area and time makes the overflow feature from OpSim3 still unnecessary in OpSim4

15. Boost and Overflow in OpSim3
# Boost factor according to the remaining observable days on sky
if self.DaysLeftToStartBoost > 0.0 and self.rankDaysLeftMax != 0.0:
observableDaysLeft = max((self.ha_twilight[fieldID]+self.ha_maxairmass) * 15.0, 0.0)
rankDaysLeft = max(1.0-observableDaysLeft/self.DaysLeftToStartBoost, 0.0)
else:
rankDaysLeft = 0.0
if self.WLtype or self.sequences[fieldID].IsActive(subseq):
(rankTime, timeWindow) = self.sequences[fieldID].RankTimeWindow(subseq,date)
rankLossRisk = max( *self.sequences[fieldID].GetRemainingAllowedMisses(subseq), 0.0)
if timeWindow:
factor = self.rankTimeMax
if self.globalProgress < 1.0:
factor = self.rankIdleSeq/(1.0-self.globalProgress)
elif self.overflowLevel > 0.0:
factor = self.rankIdleSeq/(self.overflowLevel/self.globalProgress)
rank = rankTime*factor + rankLossRisk*self.rankLossRiskMax + rankDaysLeft*self.rankDaysLeftMax

16. Deep Drilling look ahead
Remaining tracking time in OpSim4
Taken into account before sending a deep drilling sequence (short term look-ahead).
Avoid reaching tracking limits in altitude, azimuth or rotator.
Ignored in OpSim3
More important in OpSim4 due to much more observations taken close to zenith tracking limit (airmass bonus)

17. Western Bias OpSim4 removes strong western bias from OpSim3
Better area coverage algorithms
New Airmass
New HA bonus

18. Serendipity changes Proposal Id based
PropId is included in the observation
Winners and Losers counted
If serendipity is allowed, both lists are searched
Coadding is now optional
Coadding values for a target in more than one proposal can be disabled

19. Rolling Cadence OpSim4 is capable of Rolling Cadence
Area selection has parameters for time dependence
New serendipity flexibility enables clear interactions between overlapping areas

20. Proposals priorities in OpSim3
Relative priority parameter for each proposal, configurable and constant in time

21. Proposals priorities in OpSim4
New self balancing mechanism, promotes smoother progress in all proposals during the length of the survey

22. Slew time in OpSim3
slew time is handled as a bonus with a fixed shape

23. Slew time in OpSim4
slew time is handled as a cost with additional parameters in to control its shape

24. Filter change cost New preemptive filter change cost in OpSim4
Dramatically reduces the average filter change rate
In addition to the slew time cost of a filter change and the rate constraint parameters available in OpSim3

25. Final target rank
Value Boost and Cost