Automated Spacecraft Scheduling The ASTER Example Ron Cohen ronald.h.cohen@jpl.nasa.gov Ground System Architectures Workshop 2002 Jet Propulsion Laboratory
The Concept Scheduling by software instead of humans Ideally suited to orbital observing missions Rationale: Increased target data return Lower cost Now operating on ASTER / Terra (EOS-AM1) R. Cohen - Automated Scheduling 2
Some Targeting Options for Earth Observing Missions Observe Everything Valuable Data Scheduling Complexity Observed Data Observe Land Mass Only Observe Targets Only Emphasize High Priority Targets Emphasize High Priority Cloud-Free Targets R. Cohen - Automated Scheduling 3
Maximum Performance Used Capability Maximum Valuable Data Return Valuable Data Maximize: Used Capability Total Capability and Valuable Data Observed Data R. Cohen - Automated Scheduling 4
Conventional Human Scheduling Select high priority targets Check the weather report Avoid clouds Calculate observation times, pointing angles, etc. Software tools Groundtrack grid charts Generate command sequences Check constraints OR always stay well within limits Observations Spacecraft Schedule R. Cohen - Automated Scheduling 5
Questions What is the workforce cost of manually creating schedules? What is the workforce cost of adjudicating between competing requesters? (and career cost?) What about all that slack time? Are we fully utilizing spacecraft capability? Manually optimizing schedules is expensive Cassini sequences require work-years R. Cohen - Automated Scheduling 6
Automated Scheduling Automatic: Target prioritization Schedule creation Constraint checking Humans are elevated to a higher level Humans set goals, software handles the details Can still joystick the spacecraft Input desired activity with high priority Compared to conventional scheduling: More optimal (maximizes capability used and valuable data) Faster Lower cost Safer R. Cohen - Automated Scheduling 7
When is Automated Scheduling Appropriate? Shorter missions, unique events Longer missions, repetitive events Human Scheduling Cost Scheduling System Development Cost Sequence Product Magnitude (quantity, size) Automated Scheduling R. Cohen - Automated Scheduling 8
The ASTER Example Advanced Spaceborne Thermal Emission and Reflection Radiometer Built by Japan MITI/JAROS Currently operating on NASA Terra (EOS-AM1) Multiple telescopes, visual through thermal infrared Polar orbit, 16-day groundtrack repeat cycle Crosstrack pointing +/- 24 deg 60 km observation swath 10 m resolution in visible channels Many observation modes and settings Observations highly constrained by data, power, and thermal limits Many competing users R. Cohen - Automated Scheduling 9
ASTER Uplink Process!"#$%"$& '$()$*+!"#$%"$& '$()$*+ Engineering Request Acquisition Request Database Scheduler New Schedule generated each day ASTER Sequence Engineering Request Other Other instrument Other instrument Other sequences instrument sequences instrument sequences sequences EOS Operations Center Integrated Spacecraft Sequence EOS-AM1 "Terra" R. Cohen - Automated Scheduling 10
ASTER Scheduling Process -.)%/0,)%&!"#$%( '01$/(/,/#+& 2)1$)(,(& '.)3& '00#$+,/+4& 56(,)* 78().93,/#+& 7::#.,$+/,/)(!.)3,)& 50<)%$") A:"/+?& 50<)%$") -./#./,/;3,/#+!"#$%&'(()((*)+, =#>+"/+?& @*34)& =3,3 R. Cohen - Automated Scheduling 11
ASTER Acquisition Requests User specifies: Polygonal target Area of Interest (AOI) Instrument mode & settings Requirements on timing, lighting, repeat obs, etc. Request is fulfilled over time Generally multiple observations AOI can be much larger than observation swath Cloudy areas automatically reobserved R. Cohen - Automated Scheduling 12
ASTER Acquisition Requests (xars) Area of Interest R. Cohen - Automated Scheduling 13
ASTER Acquisition Requests (xars) R. Cohen - Automated Scheduling 14
ASTER Acquisition Requests (xars) R. Cohen - Automated Scheduling 15
ASTER Acquisition Requests (xars) R. Cohen - Automated Scheduling 16
ASTER Acquisition Requests (xars) R. Cohen - Automated Scheduling 17
Prioritization: Priority Varies With Position, Time, Cloudiness, Etc. Example: 3 repeat observations requested & Observed 2 times & & (low priority) Never observed & (highest & priority) Observed 3 times & Observed 1 time & & (medium priority) & (zero priority Predicted cloudy (low priority) & Data maintained globally in 1 km 2 pixels in GIS R. Cohen - Automated Scheduling 18
Prioritization: Observation Swaths Potential Observation Swath on Planet Surface Time R. Cohen - Automated Scheduling 19
Prioritization Process Overlay potential observation swath onto planet surface: & & & & & R. Cohen - Automated Scheduling 20
Prioritization Process Calculate priority of areas under swath: Priority Time R. Cohen - Automated Scheduling 21
Priority of Requests is Additive Priority = Priority 1 + Priority 2 IF single observation can contribute to both requests (common constraints: mode, lighting, time, etc.) R. Cohen - Automated Scheduling 22
Scheduling Algorithm Approach 1. Create curves of priority vs. time for each potential spacecraft state State 1 Priority Time -> State 2 Priority State 3 Priority R. Cohen - Automated Scheduling 23
Scheduling Algorithm Approach 2. Create schedule with maximum integral priority State 1 Priority State 2 Priority State 3 Priority Schedule Priority State 1 State 2 State 1 State 3 State Transitions R. Cohen - Automated Scheduling 24
ASTER Priority Function Prioriity of a point on the Earth s surface: Priority = f 0 + (f 1 f 2 " f 11 ) Prioriity adjustment factor for joysticking R. Cohen - Automated Scheduling 25
ASTER Priority Function Terms Data Collection Category Engineering, Science, Individual, etc. Ground Campaign User Category Cloudiness Predicted cloudiness, at observation time, vs. users s cloud limit Urgency Remaining Revisits Remaining Observation Opportunities Remaining Area Requested Area Time Since xar Submission Pointing Control If above desired pointing usage curve, downweight observations that request pointing changes R. Cohen - Automated Scheduling 26
2 ASTER Scheduling Systems Independent developments ASTER Ground Data System Scheduler in Tokyo Generates ASTER schedule once per day Based on JPL algorithm ASTER Mission Simulator (AMS) at JPL Second generation system Intended for mission planning Generates schedules Simulates clouds Performs observations and cloud assessment Tracks successful and failed observations Mission simulations R. Cohen - Automated Scheduling 27
AMS Examples Acquisition Requests R. Cohen - Automated Scheduling 28
AMS Examples Number of Times Areas Observed R. Cohen - Automated Scheduling 29
AMS Examples Completion of Acquisition Request Categories 100 90 80 70 60 684 Local STAR 684 GM High 684 GM Medium 684 GM Low 684 Regional High 684 Regional Medium 684 Regional Low Test 684 xar Category Completion % Complete 50 40 30 20 10 0 0 1 2 3 4 5 Years R. Cohen - Automated Scheduling 30
AMS Examples Use of Pointing Resource vs. Time 25,000 Test 684 Pointing Usage 20,000 VNIR Nominal SWIR Nominal TIR Nominal / 10 VNIR Actual SWIR Actual TIR Actual / 10 Pointing Changes 15,000 10,000 5,000 0 0 1 2 3 4 5 Years R. Cohen - Automated Scheduling 31
800 700 AMS Examples Observed Data vs. Valuable Data Test 684 - Useful Scenes & Overhead Scenes With xars Overhead 600 500 Scenes Per Day 400 300 200 100 0 0 1 2 3 4 5 Years R. Cohen - Automated Scheduling 32
Potential Future Applications Other Earth-observing missions Planetary missions Mars Reconnaissance Orbiter Coordinated and distributed observations Multiple spacecraft or sensors Multiple dynamic targets R. Cohen - Automated Scheduling 33