Cognitoware.Robotics.dll
Interface ExtendedKalmanActionModel<U, X>
U: The action type.
X: The x type.
Implemented Interfaces
Summary
A Gaussian action model that can generate a Jacobian matrix at any point.
Method Summary
ConditionBy(U, X)
Inherited from Cognitoware.Robotics.StateEstimation.ActionModel
GetError(U, X)
Inherited from Cognitoware.Robotics.StateEstimation.GaussianActionModel
GetMean(U, X)
Inherited from Cognitoware.Robotics.StateEstimation.GaussianActionModel
Returns the gradient of the non-linear action model for the specified action and x.
Details
A Gaussian action model that can generate a Jacobian matrix at any point.
This action model can be non-linear.
The Jacobian matrix represents the gradient of the non-linear function at a point on the function.
This gradient linearizes the non-linear function around a local area.
Method Details
public virtual Matrix GetActionJacobian(U action, X state)
public virtual Matrix GetStateJacobian(U action, X state)
Returns the gradient of the non-linear action model for the specified action and x.
Let Xi be the ith dimension of the input x (X = [x, y, z] = [X0, X1, X2]).
Also let fi(u, x) calculate the ith dimension of the resulting x.
Hi,j = ∂fi / ∂Xj.
Parameters:
action - The action performed at the start x.
x - The start x.
Returns:
The Jacobian matrix of the non-linear action model at the specified x and action.