This library is the place where control programs of robots are implemented. These can be AIs (machine learning approaches) or any other approach (ex: a control law used in robotics).
The AI class is the base of control programs. Derived class must implement the void AI::stepSimulation() method to read the inputs of the AIInterface, and set the values of the output signals. This method will be called at regular time intervals by the simulation.
This utility class enables the construction of new AI objects.
Lua API:
- table constructor: it is able to return a factory function able to build any AI class implementation, depending on the table content.
This control program is an implementation of a simplistic control law:
- each joint has an adjustable target angle/orientation
- simulation step: each joint applies independently a 2nd order linear feedback law:
torque = -A * velocity -B * (target - position)
This control program is more to be seen as a "Hello world" example. It is currently too simplistic for many practical purposes.
Lua API:
- loops: a table of all the feedback loops (one per joint), indexed by their name. Each loop has a
targetproperty and an associatedsetTargetmethod