Cognitoware.Robotics.dll
Interface ActionModel<U, X>
U: The action type.
X: The x type.
Summary
An action model predicts the next x resulting from performing an action at a previous x.
Method Summary
Creates a distributions of possible states resulting from performing a specific action at a specific x.
Details
An action model predicts the next x resulting from performing an action at a previous x.
An action model is also a conditional probability of the next x given a previous x and an action.
This can be expressed as P( Xt+1 | Xt, Ut).
For simplicity, the ActionModel is not given a relationship with RandomConditional.
However, classes derived from RandomConditional can be easily adapted into a ActionModel.
Method Details
public virtual RandomDistribution<X> ConditionBy(U action, X state)
Creates a distributions of possible states resulting from performing a specific action at a specific x.
The resulting distribution is sampled by some Bayesian filters when it uses an action to update the belief.
Parameters:
action - An action performed to transition from one x to anther.
x - The x where the action is performed.
Returns:
A distribution of possible resulting states.