BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Comparison of different optimization methods to construct an acqui
sition plan finding the best compromise between calculation time and algor
ithm performance
DTSTART;VALUE=DATE-TIME:20181109T083000Z
DTEND;VALUE=DATE-TIME:20181109T090000Z
DTSTAMP;VALUE=DATE-TIME:20220528T101045Z
UID:indico-contribution-960-3854@indico.esa.int
DESCRIPTION:Speakers: Cécile RENAUDIE (CNES)\nThe observation of celestia
l bodies other than Earth can without any doubt benefit from Earth observa
tion satellites. This study\, conducted for the latter\, can easily be app
lied to Phobos for example\, as part of the MMX (Martian Moons Explorer) p
roject.\n\nEarth observation satellites realize several acquisitions\, whi
ch are linked together through an acquisition plan. The calculation time n
eeded to construct such a plan is limited\; therefore the method usually e
mployed is the greedy algorithm which gives a solution quickly but not opt
imal. The temporal and kinematical constraints are numerous\, making this
a high combinatorial problem. The biggest time cost is the management of k
inematics\, especially checking if the acquisitions can be linked to each
other\, that is\, calculating the minimum duration of the rallying sequenc
e “attitude maneuver + acquisition”. In order to reduce the calculatio
n time\, the duration of each attitude maneuver and each acquisition can b
e calculated approximately\, by adding an error margin to the result in or
der to ensure the feasibility of the acquisition plan. Therefore\, a balan
ce must be found between the calculation time an optimization method leave
s to the kinematics and the margin associated to the needed approximation
. Typically\, an exact resolution method which leaves less calculation tim
e for kinematics computation needs a rough approximation\, and thus a big
margin\, which deoptimizes the calculated acquisition plan.\n\nA simplifie
d simulation model is used in order to quickly evaluate the quality of a p
lan constructed with a method (for example: greedy algorithm\, branch and
bound\, stochastic greedy algorithm\, genetic algorithm\, simulated anneal
ing\, taboo search)\, as well as the needed calculation time and therefore
the time left for kinematics computation. This model is based on a time w
indows representation.\n\nThe originality is to determine these elements a
ccording to different margins then determine which model for the calculati
on of kinematics\, with the associated margin\, gives the best results. \n
\n\n\n\n\nThe results are different depending on the studied context\; the
goal is to find the method that offers the best compromise between qualit
y of the plan and calculation time left to kinematics.\n\nhttps://indico.e
sa.int/event/224/contributions/3854/
LOCATION: Single 107
URL:https://indico.esa.int/event/224/contributions/3854/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Adaptive Pareto Front Sampling Based on Parametric Sensitivity Ana
lysis in a Bi-Objective Setting
DTSTART;VALUE=DATE-TIME:20181109T090000Z
DTEND;VALUE=DATE-TIME:20181109T093000Z
DTSTAMP;VALUE=DATE-TIME:20220528T101045Z
UID:indico-contribution-960-3865@indico.esa.int
DESCRIPTION:Speakers: Arne Berger ()\nIn order to solve non-linear multiob
jective optimization problems\, one usually solves multiple scalarized sub
problems. This provides a discrete approximation of the Pareto front which
gives useful information for the decision maker who\, in praxis\, has to
select one single solution. If the desired solution is not part of the pr
ecomputed discrete approximation one needs to apply interpolation techniqu
es.\nThis contribution shows a method which uses information from parametr
ic sensitivity analysis of the scalarized subproblems in order to choose t
he stepsize between samples adaptively to obtain a better interpolation be
tween precomputed solutions. The problems are solved with the NLP solver W
ORHP which provides sensitivity information in an efficient way by reusing
the factorization of the KKT matrix of the last optimization iteration. W
e show the basic functionality of the presented method by applying it to s
everal bi-objective optimization problems. The method can also be used for
more than two objectives if one can identify the neighboring precomputed
points which are then used for interpolation.\n\nhttps://indico.esa.int/ev
ent/224/contributions/3865/
LOCATION: Single 107
URL:https://indico.esa.int/event/224/contributions/3865/
END:VEVENT
END:VCALENDAR