Ranger NBV has been outfitted with high quality state of the art
sensor, computer, and actuator systems in order to allow it to have
precise control of its attitude and position. To allow Ranger NBV to
realize the potential of these advanced mechanical and electronic
systems, a suite of precision, state of the art attitude control
algorithms were chosen. These controllers not only take advantage of
the hardware sophistication, but also allow the pilot to deal with
large scale changes to vehicle inertia, and buoyancy due to changes
in payload, or manipulator configuration.
The vehicle provides three attitude control algorithms to the pilot, each building on the previous.
The first controller reacts proportionately to the attitude and angular velocity errors with no consideration of system dynamics. This system is more sophisticated than a simple axis based linear PD controller in that the attitude error is calculated in quaternion space which gives correct results even in the face of large slew maneuvers.
The second controller adds to the first by using the dynamic model to compute the torque necessary to follow a specified trajectory. The controller needs to have a correct model of the system parameters and dynamics in order for the computed torque to exactly match the actual torque necessary to follow accurately the desired trajectory. The quaternion based PD controller is still used, but now only has to handle errors due to inaccuracies in system parameters or system dynamics.
The third controller builds on the second by allowing the system to autonomously learn the uncertain or variable system parameters by observing the response of the vehicle as it attempts to track the commanded attitudes. Conceivably, this allows the controller to start with no system knowledge, and modify its parameters to reflect system configuration, although in practice reasonable estimates of the parameters may be determined and used as initial values. One of the methods for determining initial parameter estimates for Ranger NBV was to compare the performance of the dynamic simulation to the actual vehicle and modify the parameter values until they behavior is similar (Section 4.6).
This section starts by developing the feedback error metrics used by each of the controllers. Following this each of the three control laws are described, each one building on the information from the previous subsection.