Design and Implementation of a Multiprocessor System for Position and Attitude Control of an Underwater Robotic Vehicle

TitleDesign and Implementation of a Multiprocessor System for Position and Attitude Control of an Underwater Robotic Vehicle
Publication TypeThesis
Year of Publication1990
AuthorsAtkins EM
Academic DepartmentAeronautics and Astronautics
DegreeMaster of Science
Number of Pages107
Date Published05/1990
UniversityMassachusetts Institute of Technology
CityCambridge, MA
Thesis TypeMasters
Keywordscontrols, implementation, multiprocessor, robot, underwater
Abstract

The Multimode Proximity Operations Device (MPOD) is a neutral buoyancy simulation telerobot with the capability to fly in three dimensions and dock with an underwater satellite mockup. The vehicle may be flown from onboard, underwater remote, or surface control station. MPOD electronic systems are used to control motors, issue pneumatic commands, and read sensors. An onboard multiprocessor control system has been implemented. Five parallel processors are used to communicate with MPOD hardware and the pilot, read available sensors, calculate the vehicle state, and determine the desired control outputs.

MPOD position and attitude are determined in software via an extended Kalman filter. Sensors include a 3-axis rate sensor package, pressure transducer, pendulum inclinometers, and the 3-Dimensional Acoustic Positioning System (3DAPS), a group of underwater sound emitters and receivers.

Once the vehicle state has been determined, the control computer calculates the motor commands required to implement the desired position and/or attitude changes. For attitude and position hold, the control equations are linearized. During large angle or position maneuvers, nonlinear terms are incorporated using feed-forward linearization.  This thesis describes the electronic design, sensor integration, and multiprocessor system implementation for MPOD flight control. Simulation and experimental results are presented for vehicle state calculation. The vehicle's estimate for its state vector consistently converged to within the expected error of MPOD's sensors.