1.1 Background

The underwater environment is commonly used for simulation of the weightless environment of space. Since the late Apollo and early Skylab era, primary EVA training has been accomplished through neutral buoyancy simulation. In addition to EVA training, the underwater environment may be used for training with space hardware, or even neutral buoyancy vehicles designed to simulate spacecraft. The Space Systems Laboratory (SSL) has played a significant role in the development of underwater neutral buoyancy vehicles designed for the simulation of spacecraft operations. Since its inception at MIT in 1976, the SSL has focused on space operations, emphasizing complete, realistic tasks performed in high-fidelity neutral buoyancy simulation.
In 1983, the SSL's first free-flying neutral buoyancy vehicle, the Beam Assembly Teleoperator (BAT), began operations. BAT was "flown" underwater in an open loop flight control mode by a pilot on the surface directly commanding its ducted propellers by using a pair of 3-DOF hand controllers.
In 1984, the Multi-Mode Proximity Operations Device (MPOD) became operational. Its purpose was to study the operation of an orbital maneuvering vehicle capable of being flown from an internal cockpit or from a surface control station. MPOD was designed to incorporate a closed loop flight control system.
While a vehicle that is rotationally and vertically neutral underwater behaves similarly to a vehicle in space, there are several clear differences including water drag, thruster dynamics, and perturbations due to water current. One approach for generating accurate spacecraft-like motion in the underwater environment is to cause the neutral buoyancy vehicle to track a position and attitude trajectory specified by the output of a spacecraft simulator. To this end, students at the SSL began to instrument MPOD with sensors so that it could determine its location and orientation within the neutral buoyancy environment. [10] and [12]. In 1987, Parrish [9] simulated a 6DOF sliding mode controller for MPOD, however limitations in hardware restricted the actual implementation to 2DOF (roll and pitch). The addition of a three dimensional acoustical positioning system finally gave MPOD the ability to return full position and attitude estimates. Unfortunately, the maximum update rate for this system was between .5 and 1 Hz. Shortly after the SSL moved to the University of Maryland, Churchill [2] implemented a full 6DOF regulating controller, allowing MPOD to stationkeep at a desired position and attitude. One of the major performance limitations of this system was the low update rate of the state estimate. This work was an initial step toward full simulation of space operations through closed loop control of a neutral buoyancy vehicle. However without high sensor update rates and tracking controllers the capability was limited to regulation of position and attitude.
Near the end of 1990, the SSL began work on the Ranger Neutral Buoyancy Vehicle (Ranger NBV). The vehicle design was to draw on previous experience to create a vehicle capable of full closed loop simulation of space flight operations. Some of the areas of improvement focused on enhancing sensor precision and update rate, the inclusion of more powerful processors and precision actuators. The sensor upgrades were intended to allow more accuracy and higher update rates in vehicle attitude and position estimates. The improvements in computer hardware would allow more complex control algorithms to be executed at higher update rates, along with supporting manipulator operations. Actuator enhancements were to provide significant improvement in the ability to model and control the moments and forces applied to the vehicle. Additionally, Ranger NBV included 4 manipulators which gave the vehicle the ability to approach many realistic and complex tasks. The manipulators compose a significant portion of the vehicle mass, and changes in manipulator configuration can cause significant modifications to overall vehicle flight dynamics. This fact presented an additional challenge for design of the vehicle flight control system.
It is a fortunate coincidence that at the same time that the neutral buoyancy hardware at the SSL was experiencing revolutionary change and enhancement, a similar revolution was occurring in the world of control theory. In the late 1980's Slotine and Li described an adaptive tracking controller for robot manipulators [10]. At the same time, Wie, Weiss, and Arapostathis were studying the use of the error quaternion as the error metric for attitude control [14]. In the early 1990's, Egeland and Godhavn combined these two concepts and introduced a quaternion based adaptive tracking attitude controller [4].
These new control strategies were a natural choice for the Ranger NBV. Its higher quality sensors and actuators, could generate the information and commands necessary to take advantage of capability of these new controllers. The robotic manipulators created a highly reconfigurable system which could benefit greatly from the new passivity based adaptive control schemes.