Title: 
Planning for Earth watching and observation satellites

Presenters: 
Gérard Verfaillie and Michel Lemaître, ONERA, Toulouse, France

Tutorial abstract:
Planning activities such as detection, observation, data memorization,
analysis, and downloading for Earth watching and observation satellites
is a challenging application of automated planning and scheduling
techniques.

Since the first Earth observation satellites, this task has
evolved from handmade plans to entirely automatically generated
ones. It has also evolved from plans built off-line on the ground in
mission centers under the supervision of human operators to plans
built on-line on-board each satellite. It has finally evolved from the
management of one satellite to the centralized or distributed
management of constellations of satellites. In addition, the ability
to perform on-board detection and data analysis added reactivity
requirements. On the other hand, the abilities of the new satellites
in terms of attitude agility offered more observation opportunities,
but made the planning activity far more complex. The future Earth
watching and observation satellites will be autonomous intelligent
cooperative robots. 

The objective of this tutorial is first to present all the features of
this challenging domain, then to show how planning problems can be
stated (degrees of freedom, physical constraints, user soft and hard
requirements) using frameworks such as graph theory, integer
programming, constraint programming, scheduling and planning models,
and finally to show how they can be automatically solved using various
techniques such as greedy search, local search, tree search, or dynamic
programming.
 

Tutorial outline:
 - Earth watching and observation via satellites: what it is.
 - Activity planning for such engines: the main features.
 - Planning problems modelling: how various frameworks can be used.
 - Automatic solving: how various techniques can be used, results of
experiments, competitions and real applications.
 - The future: what remains to be done.