Developing an autonomous fleet of unmanned underwater vehicles and aerial platforms for marine missions

Abstract A new generation of Internet of Underwater Things (IoUT) has been facilitating the development of a new class of IoT apps, i.e. marine apps. Of interest in this research are Unmanned Underwater Vehicles (UUVs). Our experience with Unmanned Aerial Vehicles (UAVs) thus far suggests that UUVs have the potential to become effective and efficient when supported with optimisation algorithms. To explore this, this work first puts together a school of UUVs by using acoustic communications, and Deep Learning (DL). It then introduces an aerial platform as a source of renewable energy for the school. The resulting fleet is then empowered to operate autonomously for the purpose of carrying out marine missions and detecting underwater objects using a dual cognitive brain developed using a Dolphin Optimization Algorithm (DOA) and a Support Vector Machine (SVM). The results obtained indicate that the communication link budget parameters, reliability and localization, as well as the accuracy of coordination among the school of UUVs and the aerial platform as well as initial management of underwater missions and object detection is reasonable. Key findings at 50 kHz frequency and depths ranging between 4 and 7 meters reveal: the optimal number of acoustic Multiple Input Multiple Output (MIMO) users is 18; an acceptable Received Signal Strength (RSS) average of -61.8dB with the lowest Bit Error Ratio (BER) of $$\:{1\varvec{x}10}^{-6}$$ which indicates reasonable performance for MIMO antennas; confidence scores on the health of underwater plants ranging between 91% and 96%; a synchronous autonomy accuracy ranging between 0.91 and 0.99; an object detection accuracy ranging between 0.85 and 0.93; the supply of energy exceeds demand. Finally, a sensitivity analysis experiment on several missions does not reveal any variable outliers since changes in all input variables appear to be directly proportional to model outputs, therefore, indicating an otherwise robust model.