Description
I produced the first version of this application, named
'Parameters Tuning Unity' (PTU), during my research work about the Parameters adaptation into reactive multi-agents systems: application in mobile robotic, which it is interesting to consult.
The application's aim is to provide a support resolving issues within
parameters of an expert system or a multi-agents one (which might be expendable to other types).
Which implies:
- a client/server architecture to communicate within the system in order to finetune parameters,
- an extensible 'range' of optimization methods (such as gradient's descent).
Basically, the principle is relatively simple and can be regarded as the following protocol:
1 - creation of an XML file defining the parameters characteristics to adapt and stop criterion,
2 - PTU initialization;
- call i the index of the 'current' parameter, initialized with 0,
- initialization of the parameters with their initial value.
3 - system initialization,
4 - the PTU sends the parameters current value, to the system,
5 - the system "lives a cycle", then it sends a state evaluation (which is the one to optimize) to the PTU,
6 - the PTU uses the optimization's method selected by the user, then updates the value of the parameter i,
7 - update of i itself,
8 - the PTU checks the stop's criterion,
- if it is not reached, it returns to 4,
- or else ends the optimization's process.
9 - results' analysis.
The optimization's methods are implemented as modules, and can thus be easily added or removed.
The PTU's evolution consists mainly on the addition of such modules.
Installation
Consult the 'README' file provided with the source code.
