Abstract
Nowadays, underwater environments including ocean and karst system with their biodiversity are needed to discover and underwater vehicles are means to do that. An underwater vehicle can normally be able to classify into an under-actuated system, a fully-actuated system, or an over-actuated one, also called redundant system. This depends on the number of actuators and the number of controllable DoFs. This property decides other relating problems such as controller design, kinds of missions, building cost. Although a redundant system has drawbacks, i.e., high building cost, not easy to control, it also has a lot of merits : a good choice for fault-tolerance control, be more flexible in operations. Moreover, with redundancy, it is obvious to vary configuration in order to inherit advantages of each kind of configurations. For application speaking, underwater vehicles have a wide range of employment from civil to military such as seabed investigation, biological discovery, marine rescue, and coast monitoring. Among of them, karst exploration is a potential domain and needs to study more because of the diversity of interior ecological systems and vital role in supplying and maintaining sweet water resource, especially in European countries.Motivated by redundant feature and karst exploration, this work focuses on redundant systems in underwater robot field for exploring karst networks. Underwater vehicles with static configuration and reconfigurable one were studied. In particular, a method was proposed to determine the static configuration of a robot (position and direction of thrusters) in order to optimize performance indices : manipulability, workspace, energy, reactive, and robustness indices. An optimal solution based on multiobjective optimization was found. Simulation and experimental results were shown to prove the proposed approach. Considering advantages of a dynamically reconfigurable robot, a prototype robot, called Umbrella Robot, was built. A locally optimal approach with respect to energy-like criterion with dynamic configuration was suggested. In the meanwhile, control allocation methods for this kind of systems were investigated. Simulation results for two cases, simple guidance in which desired control vector is given and complex one in which desired control vector is able to vary in each time step, were illustrated. Our simulation and experiments validated the proposed method.