Work on the SCPLS library
Domain
Artificial Intelligence
Context
Training period for research's works initiation, in holidays' period
Period
June 2003 - September 2003
State
Finished
Language(s)
C++, Java
Description
I have performed this work during a training period, in the summer of 2003,
within the MAIA team, supervised by Laurent Jeanpierre and Alexis Scheuer.
Subject: (It's recommended to read the Scientific context before.)
The aim of the work initially consisted in producing a java version of the SCPLS library
(Sub-optimal Continuous-curvature Planning Library Set) created by Alexis Scheuer whose original version
was written in C++.
I then integrated this new version of the SCPLS library in the shape of a new module within the ToolBox
application created by Laurent Jeanpierre.
Prospectively, we can consider the implementation of this work on the Simplet which is a robot created by
Laurent Jeanpierre, Alain Dutech and Franck Gechter.
Remark(s)
The SCPLS library is always under improvement (Contact Alexis Scheuer to have information).
Installation
Scientific context
In the MAIA team, it is developed a java toolbox allowing the
quick prototyping of intelligent agents. This application aims at sharing the work of each one in order to
avoid the systematic programming of all the usual algorithms by each researcher.
The subjacent library of algorithms is based on the research works of Laurent Jeanpierre, and is mainly focused on the
Partially Observable Markov Decision Process (POMDP).
Based on an architecture with objects communicating via data flows, the interface of visual's development makes
it possible to connect modules of various origins in order to integrate them in a more complex application.
These modules thus encapsulate the algorithms of the library allowing to configure them
simply, and connecting them to other modules of the application.
In addition, Alexis Scheuer carried out in C++ a planning's library (SCPLS) which allows to calculate an optimal path
connecting two positions of the plan by respecting constraints of continuous curves.
These paths thus make it possible a vehicle to follow the plan calculated with a great reliability,
without requiring the vehicle's stop in each point of discontinuity.
profile.