High Dexterity Docking of an UUV by Fast Determination of the Area Manipulability Measure of the Arm Using ANN

The scope of this paper is to contribute towards the advancements in the autonomy and in the optimal planning of the intervention mission of an Unmanned Underwater Vehicle (UUV). Particularly, an approach is proposed for the automatic and fast determination of a high dexterity docking location. A feed forward back propagation Artificial Neural Network (ANN) is trained to calculate directly a dexterity index for the manipulator of the UUV defined in the task area and called Area Manipulability Measure (AMM). The main advantage of this procedure is the very fast determination of the AMM using the ANN compared with the very long analytical calculation of the AMM, which is used only for the training.A common intervention mission is presented and a Genetic Algorithm (GA) is implemented to search for the optimum docking pose taking into consideration the Area Manipulability Measure derived by the ANN, the distance between the current pose of the vehicle and the docking target, as well as the geometric restrictions imposed by the layout of the underwater installation.