Control Allocation for Underwater Snake Robots using Optimization Methods

摘要

Although the concept of underwater vehicles has existed for a very long time, the technology\nwithin the field has progressed substantially over the recent decades. As more and more oil and\ngas installations and operations are performed subsea, the need for more developed underwater\nvehicles is still present. A new concept which has the potential to fulfill these demands is the\nUnderwater Underwater Snake Robot. The fact that the robot is shaped as a snake makes it\nideal for moving in high viscosity environments such as water. In addition, the robot s ability\nto alter configuration provides a large workspace, and its slender and articulate body allows the\nrobot to access narrow spaces. The fact that the underwater snake robot is, itself, a manipulator\narm capable of performing light intervention tasks, makes it a powerful tool.\n\nIn this thesis, methods for thrust allocation for underwater snake robots are evaluated. Two\niterative methods, using linear and quadratic programming, are presented, developed and implemented.\nThis is also done for an explicit method for constrained thrust allocation, using redistributed\npseudo-inverse. The methods are implemented into an existing underwater snake\nrobot simulation model in Matlab/SIMULINK. Simulations are performed for unconstrained\nand constrained thrust allocation, simulating planar and thee-dimensional motion. In the unconstrained\ncase, simulations are also performed using a pre-implemented standard damped\ninverse algorithm. This is done in order to compare thrust allocation algorithm performances.\n\nIt is found that all developed algorithms produce satisfactory simulation results, although\nsome variations in performance is found. The linear programming algorithm produces a small\nerror between commanded and actual thrust, but tends to favor using a low amount of thrusters,\nwhich is not ideal. In the constrained case, the performance of this algorithm is better.\n\nThe redistributed pseudo-inverse algorithm produces a large error compared to the other\nmethods. The performance is therefore found to be sub-optimal. The quadratic programming\nalgorithm produces low errors for all simulation cases. The algorithm also tends to distribute\nthe commanded thrust more evenly amongst the thrusters. This is a significant up-side as it\nreduces wear and tear on the thrusters.\n\nThus, it is concluded that the quadratic programming algorithm for thrust allocation produces\nthe most satisfactory results, although all thrust allocation methods have proven to be\nviable for use on underwater snake robots.