The effectiveness of subsea intervention has been found to be dependent upon the capability of an Autonomous Underwater Vehicle's (AUV's) or Remotely Operated Underwater Vehicle's (ROV's) autopositioning system. However, these vessels' dynamics vary considerably with operating condition, are strongly coupled, and are expensive and difficult to derive, theoretically or by conventional testing, making the design of conventional autopilots difficult to achieve. Multi-inputhultioutput self-tuning controllers are a possible solution. Two such schemes are presented; the first is an implicit linear quadratic on-line self-tuning controller, and the other uses a robust control law based on a first-order approximation of the open-loop dynamics and on-line recursive identification. The performance of these controllers is evaluated by examining their behavior when controlling a comprehensive nonlinear simulation of an ROV and its navigation system. An interesting off-shoot of this study is the application of recursive system identification techniques to the derivation of ROV models from trials data; the potential advantages of this method are discussed.

No abstract

No abstract

Dynamic models for Remotely Operated Underwater Vehicles (ROVs) are essential for the creation of autopilots and pilot training simulators. However the derivation of these models is difficult because ROV dynamics are strongly coupled, highly nonlinear and ill-defined. In addition, expensive, specialised testing equipment is required for conventional modelling.An alternative approach which offers the potential for inexpensive and accurate ROV models uses data gathered during simple free-running trials, processed by System Identification (SI) and Parameter Estimation (PE) algorithms. SI and PE algorithms are tested on simulated ROV trials data and compared using time domain results and statistical methods.

The paper outlines the basis of Genetic Algorithm (GA) optimisation and discusses the areas where these powerful computer based search procedures can best be used. Applications include many kinds of engineering design, financial optimisation, scheduling and finding rules to describe data. An interface between a GA package and the design engineer's problem specification has been developed. The system is simple yet powerful, permitting the engineer to use sophisticated software on a PC without becoming an optimisation expert. Real applications to TLP column design optimisation and scheduling are presented to illustrate the power and flexibility of the method and the simplicity of the interface. Future additional application areas are suggested. Introduction Genetic Algorithms (GA) are a computer based search procedure which can be used for a wide range of optimisation and search problems, including those which are very difficult to handle by more conventional techniques. GAs are particularly suited to problems with nonlinearities, discontinuities or integer parameters, all of which can cause traditional optimisation schemes to fail. Applications include many kinds of engineering design and financial optimisation, scheduling and finding rules to describe data. Suitable problems are those with a solution which can be described by a set of numbers and evaluated by a computer programme to give a single measure of 'goodness'. The GA proposes a population of solutions, choosing parameters from user specified ranges, evaluates the solutions and combines the best ones to generate more and usually better candidates. Most of the computational cost usually lies in the evaluation of the worth of solutions. Typically 2000 evaluations are required for a small problem with 5 independent variables and more for larger problems. 2 Genetic Algorithms 2.1 What is a genetic algorithm? It is a fast computer based SEARCH technique, based on the Darwinian theory of Evolution (ref.1). A population of solutions evolves towards a user defined optimum. GAs find GOOD (but not necessarily optimal) solutions to any problem described by a moderate number of parameters. A major advantage is that they handle DISCRETE parameter values such as number of wells, number of casing types etc. Typical applications in the oil business include: Optimise TLP structures (cost+ weight) Optimise road/drill pad layout Optimise casing program Schedule production Find rules from data GAs perform a fast and efficient (but not necessarily perfect) search among a potentially huge number of solutions. P. 63^