![Page 1: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/1.jpg)
From Scheduling to Planning with Timelines: An history of successful Applications in Space
Part-1
Nicola Policella ESA/ESOC Germany
ICAPS 2013 Summer School Special track on P&S Space Applications Perugia, 4 June 2013
![Page 2: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/2.jpg)
From Scheduling to Planning with Timelines:
An history of successful Applications in Space
• Basic AI Planning and Scheduling concepts – successful applications in the space domain
• Project scheduling
– Precedence Constraint Posting • A common knowledge representation core based on
timelines • Timeline-based modeling and planning
• Special Track on Space Applications
2
![Page 3: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/3.jpg)
This Talk
• What is Scheduling ?
• Project Scheduling Problem – Precedence Constraint Posting – Scheduling Under uncertainty
• Final conclusions
3
![Page 4: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/4.jpg)
What is Scheduling?
• To schedule (from Merriam-Webster): 1. to appoint, assign, or designate for a fixed time 2. to place in a schedule 3. to make a schedule of
• Scheduling can generally be described as allocating (and optimizing) scarce resources over time (to perform a set of tasks).
4
![Page 5: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/5.jpg)
What is Scheduling?
• Classic view: – Scheduling is a puzzle solving activity-
• Given problem constraints and objective criterion, figure out how to best tile the capacity over time surface with operations
– Research agenda - specify new problems and/or provide new best solutions
OP1,1 OP1,2 OP1,3
OP2,1 OP2,2
R1 R2 rd1 dd1
dd2 rd2
i j
st(i) + p(i) ≤ st(j), where p(i) is the processing time of op i
i j R
st(i) + p(i) ≤ st(j) V st(j) + p(j) ≤ st(i)
rd(j) ≤ st(i) for each op i of job j
Minimize ∑ |c(j) - dd(j)|
OP1,1 OP2,2 OP1,3
OP1,2 OP2,1
R2
R1
5
![Page 6: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/6.jpg)
Scheduling vs. Planning
• In classical scheduling, all the required operations (e.g., flights, production jobs) are known at the beginning of the solving process
• In some types of problem, we may choose not to schedule all operations but typically assume that we never add to the set of operations during search
6
![Page 7: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/7.jpg)
Scheduling vs. Planning
Space operations vs. A.I. : different meanings
FACD0013|0|5|2|1|A|0| FACD0014|0|5|2|1|A|0| FACD0015|0|5|2|1|A|0| FACD0016|0|5|2|1|A|0| FACD0017|0|5|2|1|B|0| FACD0018|0|5|2|1|B|0|
A.I.
Space Operations
Planning Scheduling
Time
7
![Page 8: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/8.jpg)
Different problems • Scheduling emerges in various domains, such as
– nurse scheduling • Allocate people (with specific skills) on each shift • Regulations and working hours
– airplane landing scheduling, • Allocate services to the airplane
– production scheduling. • Series of operations • Each operation requires resources
• Different types of problems
– Dynamic scheduling • Scheduling tasks in a CPU
– Project scheduling • Organizing all the tasks for building a bridge
– Oversubscribed scheduling • Allocating satellite activities over a time interval
8
![Page 9: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/9.jpg)
Complexity ..
• Complexity of problem scheduling varies from polynomial to NP-Hard – Depends on:
• Type of constraints • Presence of resources (binary/multi-capacitive) • Optimization
• Complexity results for different classes of
scheduling problems can be found under – http://www.mathematik.uni-osnabrueck.de/research/OR/class/
9
![Page 10: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/10.jpg)
How to Live with NP-hard Scheduling Problems?
• Small sized problems can be solved by – Constraint Programming (CP)
• Complete search and inference – Mixed Integer Programming (MIP)
• Complete search and relaxation
• To solve problems of larger size one has to apply
– Approximation algorithms – Meta-heuristics / Local Search
• Incomplete search
10
![Page 11: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/11.jpg)
Project Scheduling Problem
![Page 12: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/12.jpg)
Project Scheduling (1) “.. the project scheduling problem involves the scheduling of project activities subject to precedence and/or resource constraints .. ”
• Example:
– Production scheduling – Air traffic control – Timetabling – Personnel scheduling – Railway scheduling
• A quite general scheduling problem is the Resource
Constrained Project Scheduling Problem (RCPSP)
12
![Page 13: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/13.jpg)
Project Scheduling (2)
• Scheduling process results in a (baseline) schedule
– to allocate resources to the different project
activities to optimize some measure of performance.
– basis for planning external activities, • material procurement • shipping due dates to customers
13
![Page 14: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/14.jpg)
Project Scheduling: Ingredients
• Activities
• Time (constraint)
• Resources (constraint)
14
![Page 15: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/15.jpg)
Activity
i
pi
sti eti
sti – start time eti – end time pi – processing time or duration … eti >= sti + pi wi – weight or priority
15
![Page 16: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/16.jpg)
Time
i
ri di
pi
sti eti
ri - release time ... sti>=ri di - due date … eti <= di
16
![Page 17: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/17.jpg)
Time
i
ri di
pi
sti eti
lijmin
- minimal distance (or time lag) between i and j lij
max - maximal distance (or time lag) between i and j
lij
min <= eti – stj <= lijmax
dmax – project deadline
dmax >= eti for all i
k
j
z
17
![Page 18: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/18.jpg)
Resource
i
pi
sti eti
Rk
rik
Consumable/reusable
Rk
rik
Multi-capacity
1
Binary
Rk
rik
18
![Page 19: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/19.jpg)
Project scheduling: an activity-on-the-node (AON) format
[lb, ub]
[lb, ub]
[lb, ub]
[lb, ub]
[lb, ub]
19
![Page 20: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/20.jpg)
Schedules
• A solution to a scheduling problem or schedule consists in deciding the start time (and end-time) of each activity i, such as:
– All the time constraints are satisfied (Time
Feasibility)
– All the resource constraints are satisfied (Resource Feasibility)
20
![Page 21: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/21.jpg)
Evaluate schedules • Different metrics, e.g.:
– Minimize Makespan (maximum completion time)
– Minimize Maximum Lateness
– Minimize Total (Weighted) Tardiness
makespan
i
di
i
li < 0
li > 0
li = eti - di
i
di
i
Ti = 0
Ti > 0
Ti = max{0, li}
21
![Page 22: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/22.jpg)
But …
• Is it sufficient to allocate a start (and end time) for each activity ? – Is the concept of scheduling solution sufficient ?
• Are the previous metrics the right means to
measure schedule’s quality ?
• See later …
22
![Page 23: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/23.jpg)
An example …
![Page 24: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/24.jpg)
The MARS EXPRESS Memory Dumping Problem
visibility
Overwrite!
No visibility visibility No visibility
24
![Page 25: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/25.jpg)
Relevant MARS EXPRESS parts
science housekeeping
SpaceCraft On-board memory
Communication to Earth
pack
et s
tore
s
25
![Page 26: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/26.jpg)
Example
On board activity
Channel
Already allocated in the schedule
pi
sti eti
ri di
Rk
26
![Page 27: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/27.jpg)
A Timeline Model • For each packet store (and the
communication channel) we define a different timeline
• The temporal horizon is subdivided in contiguous time intervals such that instantaneous memory operations can happen only at the edges
• Decision variables (flow values) represent the volume of data dumped within each interval
• Two types of constraints: – the ones imposed by the channel bandwidth
and – the ones by the packet stores capacity
pk1
pk2
pk3
pk4
pk5
pk6
channel
≤ packet store capacity
≤ channel bandwidth time
27
![Page 28: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/28.jpg)
t
s
Using a flow-network on the model
download
upload
initial store
pack
et st
ore c
apac
ity
channel capacity
≤
≤
Time Interval Tj
Total download
residual store
28
![Page 29: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/29.jpg)
Example 3 (3) 3 (3)
3 (3)
3 (8)
1 (10)
2 (2) 6 (6)
1 (8)
0 (10)
2 (2)
0 (2) 5 (6)
1(6)
channel
HRSC
MARSIS
T1 T2 T3 T4 T5
source
sink 29
![Page 30: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/30.jpg)
Core Algorithm • Based on max-flow
– Complete algorithm: the dumping problem has a solution iff the maximal flow over the problem horizon is equal to total data stored
– Low polynomial cost: best known O(n2.5)
• Housekeeping needs daily download – Backtracking search to
accommodate it before planning for science data
30
![Page 31: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/31.jpg)
Classic approach vs. real needs
Initial solution • Solutions which optimize
the – Makespan – Returning time: min(eti – ri)
• Very fractioned solutions
– Fractioned download
What the users wanted • Always generate a solution
(relaxation) • Control solution length • Optimize the robustness
instead ..
31
![Page 32: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/32.jpg)
Final architecture: end-to-end cycle
Parsing
Out Generation
Domain model Solver
Interaction module
32
![Page 33: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/33.jpg)
Results: users perspective • It solves the problem …
– very important precondition to be taken seriously
• Reduction of working time – Previous practice was a nightmare (every single day) – Internal evaluation estimate to 50% the reduction of
workload
• Increase of science return – Running MEXAR2 in advance identifies bottlenecks – Enable feedback to science payload PIs
• Reduction of costs
– Save commitment (i.e., money) on ground stations
33
![Page 34: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/34.jpg)
Questions ??
![Page 35: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/35.jpg)
Precedence Constraint Posting
![Page 36: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/36.jpg)
RCPSP/max • A set of activities V = {a1, a2, … , an}
– Each activity has a fixed processing time, or duration, pi. – Any given activity must be scheduled without preemption.
• A set of Temporal Constraints designate minimum and maximum time lags between the start
times of two activities – lij
min <= sj – si <= lijmax
• A set of Reusable Resources. During their processing, activities require specific resource
units from a set R = {r1 , … , rm} – Resources are reusable, i.e. they are released when no longer required by an activity and are then
available for use by another activity – Each activity ai requires of the use of reqik units of the resource rk during its processing time pi. – Each resource rk has a limited capacity of ck units. – For each resource:
• ∑ 𝑟𝑟𝑟𝑖𝑖 ≤ 𝑐𝑖𝑖𝑐𝑖≤𝑡<𝑒𝑖
• A schedule is an assignment of start times to activities a1, a2, … , an, i.e., S = (s1, s2, … , sn)
36
![Page 37: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/37.jpg)
Precedence Constraint Posting
• A solving paradigm based – Constraint programming
• CSP
– The presence of Simple Temporal Problem • Used to identify and analyze the current solution
– Iterative Repair
• Precedence Constraints are posted to repair the current solution
37
![Page 38: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/38.jpg)
Constraint Satisfaction Problem (CSP)
• An instance of CSP involves – a set of Decision Variables X= {X1, X2, …., Xn}
– a Domain of possible values Di for each variable
– a set of Constraints C= {C1, C2, …, Cq}, such that Cj ⊆ D1 × D2 × … × Dn
• A Solution is an assignment of domain values to all variables consistent with all the constraints Cj
38
![Page 39: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/39.jpg)
Solving Approach
• Based on two aspects: – Propagation of constraints – Decision of new constraints to solve current
conflicts (next slides)
Problem
Propagation
Constraint posting
Solution
39
![Page 40: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/40.jpg)
Solving Approach
Solution !
40
![Page 41: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/41.jpg)
Simple Temporal Problem (STP) • A Simple Temporal Problem is a set of n variables (time points) {tpi} with
domain [lbi,ubi] and a set of constraints {a≤ tpj− tpi≤ b}.
• A particular time point tp0 called time origin with domain [0,0]
• The problem is consistent when an instantiation of the variables {tpi} exists such that satisfies all the constraints
• A time-map represents a Simple Temporal Problem.
[0,0] tpi
[lbi, ubi] time map
tp0
41
![Page 42: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/42.jpg)
An example of time-map
[60,100]
[30,40]
[10,20] [10,50]
[20,30]
[0,0]
[10,50] [30,70]
[10,20] [20,70] [30,90]
42
![Page 43: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/43.jpg)
STP Properties • STP is a special case of Constraint Satisfaction Problem (CSP)
• Can be reduced to a Shortest Path Problem on a distance graph Gd
– Inconsistency <=> negative cycles in Gd – Variables (tpi) domains <=> Shortest Path Trees
• Supports incremental modification
– Insertion of new constraints – removal of constraints
• Has an equivalent set of constraints called Minimal Network
43
![Page 44: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/44.jpg)
STP as a Shortest Paths Problem • STP problem can be reduced to a shortest paths problem
on a graph Gd(V,E), where V is the set of time points and E is the set of labelled edges such that:
[a,b] tpi tpj +b
-a
tpi tpj
• A STP is consistent iff Gd has no negative cycles [Decter et al, 1991].
• An STP has at least two solutions: • EST, in which each variable is assigned to its earliest
possible time • LST, in which each variable is assigned to its latest
possible time
44
![Page 45: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/45.jpg)
Precedence Constraint Posting (PCP)
• Variables: – start and end times of each activity
• Domains – A schedule horizon [0, H]
• Constraints – Temporal constraints (e.g., duration of activities, min-max separation
between activities, simple precedence) – Resource constraints (e.g., bounds on capacities)
time
resource
Temporal constraint propagation
Resource profile analysis
Earliest Start Time Solution
Precedence constraint
45
![Page 46: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/46.jpg)
46
Precedence Constraint Posting (PCP)
[2, 5]
4 18
4
5
1
1 5 1 0
6 12
0
loop propagate(CSP) compute-conflicts(CSP) if no-conflict then return-solution else if unresolvable-conflicts then return-fail else select-conflict select-precedence post(precedence) end-loop
Remove resource violations posting further precedence constraint in the temporal network
cj
SOLUTION FOUND! Temporal Net Inconsistent values can be pruned in polynomial time (Dechter, 91)
Resource profile
![Page 47: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/47.jpg)
Solving Approach
• Three steps: – Identify conflicts – Conflict ordering – Conflict resolution
• Based on the works from • Chien & Smith, • Cesta, Oddi & Smith • Policella, Cesta, Oddi, & Smith
47
![Page 48: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/48.jpg)
Identify conflicts
• Analyze the Earliest Start Solution profile – Earliest start time allocation assures to have time feasible solution
• Compute Conflict Peaks
– there is a conflict peak on resource rk at time t if the resource requirement of the activities scheduled in t exceeds the resource capacity capk of rk
• Compute MCSs on Peaks
– a Minimal Critical Set (MCS) is a conflict such that each of its proper subsets is not a conflict
48
![Page 49: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/49.jpg)
MCSs? (Example)
Peak
A1
5 5 1
A2 A3 A4
Pairs of activities: {A1 A2}{A1 A3}{A1 A4} {A2 A3} {A2 A4}{A3 A4}
Resource capacity = 8
1
MCSs: {A1 A2} [Erschler et al., 1990]
49
![Page 50: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/50.jpg)
Approximate Computation of MCSs
• The number of minimal critical sets is exponential in the general case
• Proposal: Sampling them with an approximate analysis on peaks – Linear sampling – Quadratic sampling
50
![Page 51: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/51.jpg)
Two Peak Sampling Methods (Example)
Peak
A1
4 4 3 3 1 2 2 A2 A3 A4 A5 A6 A7
Linear Sampling: {A1 A2} {A2 A3} {A3 A4 A5} {A4 A5 A6}
Quadratic Sampling: {A1 A2} {A1 A3} {A1 A4} {A2 A3} {A2 A4} {A3 A4 A5}{A3 A4 A6}{A3 A4 A7} {A4 A5 A6}
Resource capacity = 6
51
![Page 52: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/52.jpg)
Conflict ordering • Which MCS to resolve first
– Use estimator K [Laborie&Ghallab,IJCAI’95] to order MCSs • Least commitment strategy [Weld, AIM 94]
– “Select the MCS that is temporally closest to an unsolvable
state”
– Constraint: ρ = 𝐴 𝑏𝑟𝑏𝑏𝑟𝑟 𝐵 • Where A – B in [dmin, dmax]
• 𝑐𝑏𝑐𝑐𝑐𝑐 𝜌 = min 𝑑𝑑𝑑𝑑 ,0 −min (𝑑𝑑𝑖𝑑 ,0)𝑑𝑑𝑑𝑑−𝑑𝑑𝑖𝑑
– Set of constraints that can be posted Φ = 𝜌𝜌, … , 𝜌𝜌
• 1𝐾(Φ)
= ∑ 11+𝑐𝑐𝑑𝑑𝑖𝑡 𝜌𝑖 −𝑐𝑐𝑑𝑑𝑖𝑡(𝜌𝑑𝑖𝑑)
𝑖𝑖=1
52
![Page 53: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/53.jpg)
Conflict resolution
• how to choose the precedence constraint – Use slack-based heuristics [Smith&Cheng,AAAI’93]
Peak
A1
5 5 1
A2 A3 A4 1
MCSs: {A1 A2} A1 before A2 OR A2 before A1 ??
53
![Page 54: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/54.jpg)
Identify a contention peak and post a leveling constraint
Activity 2
R1 Activity 1 Activity 2
R1 Activity 1
Advantages • Retain flexibility implied by problem constraints
(time and capacity) • Can establish conditions for guaranteed
executability • Mixed initiative • Be exploited in planning and scheduling
integration
Why Precedence Constraint Posting ?
54
![Page 55: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/55.jpg)
Questions ??
![Page 56: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/56.jpg)
Project Scheduling under Uncertainty
![Page 57: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/57.jpg)
What’s Missing from the Classical View of Scheduling
• Practical problems can rarely be formulated as static optimization tasks
– Ongoing iterative process
– Situated in a larger problem-solving context
– Dynamic, unpredictable environment
57
![Page 58: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/58.jpg)
58
Scheduling with Uncertainty
Resource availability
temporal uncertainty
resource uncertainty
causal uncertainty
Activities last longer than expected or they can be postponed
A new precedence relation between a pair of activities requires a revision of previous choices
Difference between nominal (left) and actual (right) resource availability. Reduction of resource availability blocks the execution of some activities an their consequent delay
![Page 59: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/59.jpg)
Managing Change
• “Scheduling” is really an ongoing process of responding to change
• Predictable, Stable Environment
•Optimized plans
• Unpredictable, Dynamic Environment
•Robust response
Manufacturing
Crisis Action Planning
Project Management
59
![Page 60: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/60.jpg)
Approaches to Managing Change
– Build schedules that retain flexibility
– Produce schedules that promote localized recovery
– Incremental re-scheduling techniques (e.g., that consider “continuity” as an objective criteria)
– Self-scheduling control systems
60
![Page 61: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/61.jpg)
Approaches to Managing Change
• Approaching a scheduling problem requires the coupling of:
1. a predictive scheduling engine, to propose a possible
solution
2. a reactive scheduling engine, to manage the current solution and to make “repairs” during the execution
61
![Page 62: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/62.jpg)
Incremental Schedule Repair
• Several competing approaches to maintaining solution stability – Minimally disruptive schedule revision (temporal delay,
resource area, etc.) – Priority-based change – Regeneration with preference for same decisions
• Even less understanding of how to trade stability
concerns off against (re)optimization needs
62
![Page 63: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/63.jpg)
Current Approaches
robust solutions [Davenport et al., 01]
partial schedule [Wu et al., 99]
local reactive [Smith, 94] global reactive [El Sakkout & Wallace, 2000]
dynamic scheduling
On-line (or Reactive)
Off
-line
(o
r Pre
dict
ive)
• Producing solutions based on uncertainty knowledge
– Synthesis of robust solutions [Davenport et al., 01]
• Using flexibility to cope with uncertainty – Partially defined solutions [Wu et al., 99]
• Dealing with uncertainty during schedules execution
– Local versus Global approach [Smith, 94; El Sakkout & Wallace, 2000]
pred
ictiv
e re
activ
e
63
![Page 64: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/64.jpg)
Current Approaches (2) ROBUSTNESS • different views:
– Execution based. • small & fast repairs [Ginsberg et al., 98]. Not dependent of the
rescheduling algorithm used. – Quality based.
• maintaining the makespan (quality) stable [Leon et al., 94] – Solution based
• maintaining the solution close to the original one [Leus et al., 04]
64
![Page 65: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/65.jpg)
Partial order schedule • Given a scheduling problem a Partial Order Schedule (POS) is a set of solutions
that can be represented as a temporal graph [any time feasible schedule defined in the graph is also a resource feasible schedule]
• An interesting property of POSs is “fast re-scheduling”: – many external changes may have a reactive response accomplished via
simple propagation in the underlying temporal network (a polynomial time calculation).
• Sequence activities that compete for resources letting start and end times float
65
time R1
OP2,1 OP1,2 OP2,1 OP1,2
time
![Page 66: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/66.jpg)
Partial order schedule
scheduler
Execution monitoring
Fixed-time solutions
scheduler
Execution monitoring
”fast” scheduler
POSs
66
![Page 67: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/67.jpg)
How to compute POSs
• Variables: – start and end times of each activity
• Domains – A schedule horizon [0, H]
• Constraints – Temporal constraints (e.g., duration of activities, min-max separation between
activities, simple precedences) – Resource constraints (e.g., bounds on capacities)
time
resource
Temporal constraint propagation
Resource profile analysis
Temporally flexible solution
Precedence constraint
67
![Page 68: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/68.jpg)
Two different resource profiles • Because of the flexibility of the temporal network an
exact computation of demand profile is not possible • An intuitive compromise consists of computing its
upper-bound and lower-bound projections
• A different way is to compute the resource profile for a specific point in the search space: – Earliest start time solution
68
![Page 69: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/69.jpg)
EBA: Envelope Based Analysis
• A least commitment POS generation
• Maintain the flexibility of temporal network
• Compute the tightest possible resource bounds of the “flexible” profile using Resource Envelope computation [Muscettola, 02]
69
![Page 70: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/70.jpg)
ESTAC: Earliest Start Time Analysis • Computing a Resource Profile
in the Earliest Start Time of the temporal network
• Two Step process – Generate a single solution
reasoning on the earliest start time profile
– Post process the solution transforming it into a POS through chaining
70
![Page 71: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/71.jpg)
in a nutshell ..
EBA ESTAC
71
![Page 72: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/72.jpg)
Scheduling with Uncertainty • Still an open problem
– Type of scheduling problem has to be considered – Type of uncertainty is crucial
• Execution environment
• For types of uncertainty [De Meyer et al., 2002]: – Variation
• combined small influences (worker sickness, weather, delayed deliveries, unanticipated difficulties of activities, resource unavailability, etc.)
– Foreseen uncertainty • influences that are well-understood but that the project management team
cannot be sure will occur (for example, side effects of new drugs) – Unforeseen uncertainty,
• Cannot be identified during project planning. Typically occur in projects that push a technology envelope or enter a new or partially known market
– Chaos
72
![Page 73: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/73.jpg)
Scheduling: Final considerations and Challenges
![Page 74: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/74.jpg)
Building scheduling applications
• Four reasons are given for the lack of success:
1. complexity of knowledge elicitation;
2. uncertainty;
3. difficulties in human–computer interaction;
4. oversimplification of the problem.
74
![Page 75: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/75.jpg)
Methodological Issues
• How relevant are the problems that are solved? – Idealized constraint models – Emphasis on easier objective criteria – Real problems do not have random structure
• Overcoming these objections is clearly one
continuing direction for scheduling research
75
![Page 76: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/76.jpg)
Challenges
• Scheduling models that incorporate richer activity models
• Can integrated P & S problems really be solved as one big optimization task?
• How to achieve tighter interleaving of action selection and resource allocation processes
• Mixed-Initiative Scheduling
• Requirement Analysis
76
![Page 77: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/77.jpg)
Challenges
• Scheduling models that incorporate richer activity models
• Can integrated P & S problems really be solved as one big optimization task?
• How to achieve tighter interleaving of action selection and resource allocation processes
• Mixed-Initiative Scheduling
• Requirement Analysis
77
![Page 78: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/78.jpg)
Activity
i
ri di
pi
sti eti
Inside the activity. Logic ? • Part of a process
• A process itself
78
![Page 79: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/79.jpg)
Another example …
![Page 80: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/80.jpg)
Mission Description • Alphasat, based on the new Alphabus platform, will be delivered to orbit to be
operated by Inmarsat in 2013. – It will carry an Inmarsat commercial communication payload – 4 Technology Demonstration Payloads (operated as secondary payloads)
• TDPs:
– An advanced Laser Communication Terminal; – A Q-V Band communications experiment; – An advanced Star Tracker with active pixel detector; – An environment effects facility to monitor the GEO radiation environment and
its effects on electronic components and sensors.
80
![Page 81: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/81.jpg)
TDPs ESA Coordination Office (TECO)
TDPs Operations Centre (s)
Inmarsat Ground
Control Centre
Planning input from/to TECO
Internet
Planning Products from/to INM-FTP & TDPs OCs
Planning Products from/to TECO
TDPs Operations Centre (s)
TDPs Operations Centre (s)
TDPs Operations Centre (s)
Workflow and planning interfaces
Main goals: • TDPs activity coordination +
planning support – Collection of activity/task
requests – Conflict identification and
resolution – Generation of the final
activity plan
• The planning process is completely automated
81
![Page 82: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/82.jpg)
Tasks & Constraints • Task Requests
• TDP • Feasible Time interval • Sequence of subtasks <dur, submode, bandwidth,
power> • “On-board only”/ “on-board + on-ground” • A weight value w
• Constraints
• Among TDPs – e.g., TDP1.B during TDP2.Z • TDP vs S/C – e.g., TDP1.F not-during S/C.manoeuvres • Among tasks – e.g., tr1<tr2 and allocate(tr1) iff
allocate(tr2) • Resource constraints – e.g., bandwidth and power
limits
B C F C
lbt ubt
d0,t d1,t d2,t d3,t
ground
on board
power
bw
TDP
82
![Page 83: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/83.jpg)
Problem and solution • Problem:
• A set of task requests • A set if initial submode for each TDP • A set of constraints • S/C status and availability
• Solution
• A set S of allocated tasks • Maximize 𝑉𝑉𝑉𝑉𝑟(𝑠) = ∑ 𝑤(𝑐𝑟)𝑡𝑟∈𝑆
83
![Page 84: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/84.jpg)
Example – problem
ground
on board
TDPb
TDPa
not available not available
not available
ta1
tb1
tb1
F A
tb2
tb2
F C B
W X X
TDPa.W DURING TDPb.A
tb1 < 10 + tb2
Z X
ta2
ta2
power
bw
84
![Page 85: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/85.jpg)
Example – solution
ground
on board
TDPb
TDPa
not available not available
not available ta1
tb1
F A
X W
tb2
tb2
F C B
X
TDPa.W DURING TDPb.A
tb1 < 10 + tb2
tb1
power
bw
85
![Page 86: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/86.jpg)
Challenges
• Scheduling models that incorporate richer activity models
• Can integrated P & S problems really be solved as one big optimization task?
• How to achieve tighter interleaving of action selection and resource allocation processes
• Mixed-Initiative Scheduling
• Requirement Analysis
86
![Page 87: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/87.jpg)
Integrating Planning & Scheduling
Mixed-Initiative Model
Waterfall Model
Plan Schedule
Planner
Scheduler
Planner
Scheduler Schedule
Plan
“Planning & scheduling are rarely separable”
87
![Page 88: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/88.jpg)
Challenges
• Scheduling models that incorporate richer activity models
• Can integrated P & S problems really be solved as one big optimization task?
• How to achieve tighter interleaving of action selection and resource allocation processes
• Mixed-Initiative Scheduling
• Requirement Analysis
88
![Page 89: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/89.jpg)
Mixed-Initiative Scheduling Challenges
• Management of user context across decision cycles
• Explanation of scheduling decisions – Why did you do this? – Why didn’t you do that?
• Adjustable autonomy
89
![Page 90: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/90.jpg)
The same example …
![Page 91: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/91.jpg)
Explanation Need of generating explanation
– TECO as the system has been designed to be completely automated – A proper explanation is also needed to have effective iterations
between TDPOCs and TECO. The approach is based on the following points: • A “protocol” to exchange information between TDP-OCs and TECO • “Labeled decisions” with information about the solver and the
motivation of the decision. • An “Explanation Generator” module to generate the information
for the system users by applying the given protocol.
91
![Page 92: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/92.jpg)
Explanation – examples
ground
TDPb
TDPa
not available
ta
tb
F A
X W X
TDPa.W DURING TDPb.A
ground
TDPa
TDPs OFF
ta
X W X X ≠ OFF
planner: B before A scheduler: no solution Explanation should consider 1) scheduling conflict 2) planner decision 3) domain theory
scheduler: A before TDPs OFF planner: no solution Explanation should consider 1) planning conflict 2) domain theory
92
![Page 93: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/93.jpg)
Explanation • The final explanation is not generated directly by the different solvers
– the solvers can have only a limited view of the current situation – to allow decoupling the set of used solvers from the final explanation
generation process (and the associated protocol)
• The approach consists in “tracing back” all the decisions – collecting the different annotations added during the solving process. – identifying the specific case based both on the solvers and on the content of
the different annotations
• Modular approach to facilitate future re-usability and evolution – a different type of user accessing the TECO system (e.g., web-client) which can
require a different protocol – modifying the set of solvers.
93
![Page 94: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/94.jpg)
Challenges
• Scheduling models that incorporate richer activity models
• Can integrated P & S problems really be solved as one big optimization task?
• How to achieve tighter interleaving of action selection and resource allocation processes
• Mixed-Initiative Scheduling
• Requirement Analysis
94
![Page 95: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/95.jpg)
Requirements Analysis • “Scheduling is really a process of getting the constraints right”
• Current tools designed around a “Specify and Solve” model of user/system
interaction – Inefficient problem solving cycle
• Optimization techniques try to solve the problem WHILE a human changes
the problem to make it solvable!
• Human scheduler decisions are often based on knowledge not represented in the scheduling problem!
• Mixed-Initiative solution models – Incremental solution of relaxed problems – Iterative adjustment of problem constraints, preferences, priorities
95
![Page 96: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/96.jpg)
Research Directions for the Next 10 Years
• Deeper integration of AI and OR techniques
• Robust schedules and scheduling
• Global coherence through local interaction
• Extension to larger-scoped problem-solving processes
• Rapid construction of high performance scheduling services
96
![Page 97: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/97.jpg)
Thank you !
![Page 98: Icaps 2013 - From Scheduling to Planning with Timelines: An ...icaps13.icaps-conference.org/wp-content/uploads/2012/06/...ICAPS 2013 Summer School Special track on P&S Space Applications](https://reader034.vdocuments.us/reader034/viewer/2022052409/608c3c48079b1c142f693ed3/html5/thumbnails/98.jpg)
References • P. Brucker, A. Drexl, R. Mohring, K. Neumann, and E. Pesch, “Resource-Constrained Project Scheduling: Notation, Classification, Models, and Methods,”
European Journal of Operations Research. 1998 • A. Cesta, A. Oddi, and S.F. Smith, “A Constraint-Based Method for Project Scheduling with Time Windows,” Journal of Heuristics, n. 8, pp. 109–136, 2002 • A. Cesta, G. Cortellessa, M. Denis, A. Donati, S. Fratini, A. Oddi, N. Policella, E. Rabenau, and J. Schulster, ”MEXAR2: AI Solves Mission Planner Problems”, in
IEEE Intelligent Systems, pp. 12-19, issue July/August 2007. • C. Cheng, and S.F. Smith. “Generating Feasible Schedules under Complex Metric Constraints.” In Proceedings 12th National Conference on AI (AAAI-94), 1994 • A. J. Davenport, C. Gefflot, and J. C. Beck. Slack-based Techniques for Robust Schedules. In Proceedings of 6th European Conference on Planning, ECP-01, 2001. • A. De Meyer, C. H. Loch, and M. Tisch, “Managing project uncertainty: From variation to chaos,” MIT Sloan Management Review, pp. 60–67, Winter 2002. • R. Dechter, I. Meiri, and J. Pearl. “Temporal Constraint Networks.” Artificial Intelligence 49, 61–95, 1991 • H. H. El Sakkout and M. G. Wallace. Probe Backtrack Search for Minimal Perturbation in Dynamic Scheduling. Constraints, 5(4):359–388, 2000. • J. Erschler, P. Lopez, and C. Thuriot. Temporal reasoning under resource constraints: Application to task scheduling. In Advances in Support Systems Research,
1990 • M. L. Ginsberg, A. J. Parkes, and A. Roy. Supermodels and Robustness. In Proceedings of the 15th National Conference on Artificial Intelligence, AAAI-98, pages
334–339. AAAI Press, 1998. • P. Laborie and M. Ghallab, “Planning with Sharable Resource Constraints.” IJCAI 1995 • R. Leus. The generation of stable project plans. PhD thesis, Katholieke Universiteit Leuven, 2003. • V. Leon, S. D. Wu, and R. H. Storer. Robustness measures and robust scheduling for job shops. IIE Transactions, 26(5):32–43, September 1994. • N. Muscettola. “Computing the Envelope for Stepwise-Constant Resource Allocations”. In Principles and Practice of Constraint Programming, 8th International • Conference, CP 2002, 2002. • A. Oddi and N. Policella, “Improving Robustness of Spacecraft Downlink Schedules”, in IEEE Transactions on Systems, Man, and Cybernetics (Part C), vol. 37, no.
5, pp. 887-896, 2007. • N. Policella, “Scheduling with uncertainty — A proactive approach using partial order schedules,” Ph.D. Thesis, Italy: Università degli Studi di Roma “La
Sapienza”, 2005. • N. Policella, A. Cesta, A. Oddi, and S.F. Smith, “From Precedence Constraint Posting to Partial Order Schedules: A CSP approach to Robust Scheduling”, in AI
Communications, vol. 20, no. 3, pp. 163-180, 2007. • N. Policella, A. Cesta, A. Oddi, and S.F. Smith, “Solve-and-robustify: Synthesizing partial order schedules by chaining”, in Journal of Scheduling, 2008 • N. Policella, H. Oliveira, and E. Benzi, “Planning Spacecraft Activities: An automated approach”, in ICAPS 2013 • S.F. Smith and C. Cheng, “Slack-Based Heuristics for Constraint Satisfaction Scheduling.” In Proceedings 11th National Conference on AI (AAA1-93), 1993 • S. F. Smith. Reactive Scheduling Systems. In D. Brown and W. Scherer, editors, Intelligent Scheduling Systems. Kluwer Academic Publishers, 1994. • D. Weld, An Introduction to Least Commitment Planning. In AI Magazine, 1994 • S. D. Wu, E. S. Beyon, and R. H. Storer. A graph-theoretic decomposition of the job shop scheduling problem to achieve scheduling robustness. Operations
Research, 47(1):113–124, January – February 1999. 98