Description: Primitive AI Interceptor Controller test.
Combined motion and rotation controller with calculated direction for predictive aim.
The early version of the AI flight controller prototype without a decision module.
The interceptor chooses the nearest target among the available ones.

Description: LQR Controller test.
Single controller test using variable additive white gaussian noise on a sinusoidal signal.
Scalar and discrete-time version of the controller with infinite horizon.
Signal generation : 60 Hz

Description: Primitive Motion Controller test
The next version of the primitive motion controller used to intercept the object.
The red object uses a motion controller.
The orange object uses a guidance method with direct motion data.
The diagram containing the controller description is included at the beginning of the presentation.

Description: Primitive trackball turret controller test.
The first version of the prototype rotation controller of the trackball turret.
The presentation shows two automatic turrets placed on the rotating object.
Turrets are trying to calculate the optimal direction in order to launch a projectile.
The diagram containing the controller description is included at the beginning of the presentation.

Description: Primitive Motion Controller test
Modification of the object's interception procedure by adding a PI controller.
The simulation shows the difference between the old and the new method.
New method:
- Filtering target position vector
- Calculation of target data using derivation of equations of motion
- Prediction of target data using Dead Reckoning
- Calculation of the recommended acceleration vector by one of the
guidance methods.
Old method:
- Filtering target position vector
- Calculation of target data using derivation of equations of motion
- Prediction of target data using Dead Reckoning
- Calculation of the recommended acceleration vector by one of the
guidance methods.
- Calculating the PI controller's output value.

Description: PI Controller test.
Single controller test using variable additive white gaussian noise on a sinusoidal signal.
Selected one of the methods of tuning the controller parameters.
Standard test of the controller in a continuous version where:
Error = SetPoint-ProcessVariable
Signal generation : 60 Hz

Description: Recursive Least Squares Filter test.
Single adaptive filter tests using variable additive white gaussian noise on a sinusoidal signal.
Adaptive filter is in system identification mode.
Faster version of the LMS filter.
Visible faster adaptation at the end of the test.
The size of the buffer : 50-60
Signal generation : 60 Hz

Description: Least Mean Squares Filter test.
Single adaptive filter tests using variable additive white gaussian noise on a sinusoidal signal.
Adaptive filter is in system identification mode.
Calculated in real-time optimal learning rate.
Visible slow adaptation at the end of the test.
The size of the buffer : 50-60
Signal generation : 60 Hz

Description: Automatic control of the physics-based trackball turret with Predictive aiming.
Angular version of the Cosine rule was used to align one of the axis of the turret with the target vector.
Target vector calculated using a different form of Cosine rule.
The turret returns to the starting orientation when there is no target in its range.
The turret fires projectile two times a second.
Continuous collision detection between projectiles and target.
The two displays at the bottom show the number of hits and precision.
The number of hits and the precision of the turret : 441/99%
Quaternion form of RK2/RK4 integration used for modify orientation, angular velocity and angular acceleration of the trackball turret.

Description: Automatic control of the physics-based trackball turret with LOS aiming.
Angular version of the Cosine rule was used to align one of the axis of the turret with the target vector.
Target vector calculated using the Line-of-Sight method.
The turret returns to the starting orientation when there is no target in its range.
The turret fires projectile two times a second.
Continuous collision detection between projectiles and target.
The two displays at the bottom show the number of hits and precision.
The number of hits and the precision of the turret : 0/0%
Quaternion form of RK2/RK4 integration used for modify orientation, angular velocity and angular acceleration of the trackball turret.

Description: The method of interception has been modified by filtration of the object's position vector.
Direct input of the object position.
The simulation shows the difference after modifying the method.
Precision dependent on the filtration method.
Current method:
- Calculation of target data using derivation of equations of motion
- Prediction of target data using Dead Reckoning
New method:
- Filtering target position vector
- Calculation of target data using derivation of equations of motion
- Prediction of target data using Dead Reckoning

Description: The method of interception has been modified by filtration of the object's acceleration vector.
Acceleration of the object calculated from a derivative.
The simulation shows the difference after modifying the method.
Precision dependent on the filtration method.
Current method:
- Calculation of target data using derivation of equations of motion
- Prediction of target data using Dead Reckoning
New method:
- Calculation of target data using derivation of equations of motion
- Filtering target acceleration vector
- Prediction of target data using Dead Reckoning

Description: A comparison of two methods that determine current target data such as position, speed and acceleration.
The observable effect of these methods on pursuers during a standard test.
First method:
- Direct input
Second method:
- Calculation of target data using derivation of equations of motion
- Prediction of target data using Dead Reckoning

Description: Angular version of the Cosine rule method.
The method calculates the optimal time and recommended direction of the acceleration axis to intercept the target direction.
This method calculates the optimal time using a 4th degree polynomial.
Near-optimal "angular intercept" case.
The allocation of the random orientation and angular velocity for all tested objects during initialization.
First test :
Interception of the destination vector by one of the axis of the object.
Quaternion form of RK2/RK4 integration used for modify orientation, angular velocity and angular acceleration of the tested objects.

Description: Orientation matching using Cosine Rule test.
Near-optimal "orientation matching" case.
The allocation of the random orientation and angular velocity for all tested objects during initialization.
First test :
Adjusting the direction of one of the object's axes to the direction of the target vector.
Second test :
The equalization of the orientation of the source object to the orientation of the target object.
Quaternion form of RK2/RK4 integration used for modify orientation, angular velocity and angular acceleration of the tested objects.

Description: Angular version of the LOS guidance.
Suboptimal "angular intercept" case.
The allocation of the random orientation and angular velocity for all tested objects during initialization.
First test :
Interception of the destination vector by one of the axis of the object.
Second test :
Interception of all axes of the target object by the axes of the source object.
Quaternion form of RK2/RK4 integration used for modify orientation, angular velocity and angular acceleration of the tested objects.

Description: Self-tuning Dahlin PID Controller test.
Multiple self-tuning controller tests using variable additive white gaussian noise on a sinusoidal signal.
Only for this test :
Set Point = Generated input signal
Process variable = Controller output
Signal generation : 60 Hz
K factor : 1.0-0.0
Three other factors : 0.1

Description: Optimized augmented proportional navigation test.
The method uses additional information about pursuer acceleration.
Near-optimal "interception" case.
Two moving pursuers and evader.
Acceleration of pursuers 1/3 greather than acceleartion
of evader.
The interception of the target takes place as quickly as possible.

Description: The method calculates the optimal time and recommended direction of the acceleration vector to the closest approach to the target.
The equation of this method can be obtained from the first derivative of another equation used for continuous collision detection.
This method can be considered as "higher-order" proportional navigation.
Near-optimal "closest approach" case.
Two moving pursuers and evader.
Acceleration of pursuers 1/3 greather than acceleartion
of evader.
The interception of the target takes place as quickly as possible.

Description: The method calculates the optimal time and recommended direction of the acceleration vector to intercept the target.
This method calculates the optimal time using a 4th degree polynomial.
Near-optimal "intercept" case.
Two moving pursuers and evader.
Acceleration of pursuers tripple greather than acceleartion
of evader.
The interception of the target takes place as quickly as possible.

Description: Kalman filter test.
Multiple filter tests using variable additive white gaussian noise on a sinusoidal signal.
6-level Kalman filter adapted to motion tracking, smoothing and prediction.
A defined 6x6 Transition matrix to achieve the next state (Taking into account Position-Velocity-Acceleration-Jerk-Jounce-Crackle).
The control vector and the input matrix were not used in this case.
The value of measurement noise depends on the variance.
Implementation with a reduced amount of matrix in order to:
-Eliminate redundant multiplications by 0 or 1
-Acceleration of the process of calculating optimal gains
-Observe one state (matrix inversion reduced to 1/value)
-Optimal gains are recalculated only when the value of key
variables changes.
Signal generation : 60 Hz
Process noise variable : 1.0-0.0

Description: The method calculates the optimal time and recommended direction of the acceleration vector to join the formation.
Near-optimal "joint the formation" case.
Two moving wingmen and leader.
Acceleration of wingmen tripple greather than acceleartion
of leader.
Formation is always created as soon as possible.