Cooperative autonomous marine vehicles for adaptive passive acoustic monitoring

该研究项目旨在开发一群能够对近岸环境进行声学监测的自主海上车辆。这将包括机器人系统的硬件设计和智能软件的实现，使车辆能够以合作的方式执行任务。机器人车辆将利用水听器传感器被动监测海洋环境，以便探测和定位海面以下的单个噪声源（即海洋哺乳动物，潜艇车辆，水下涡轮机等）。最初，将根据需要交换的信息量以及机器人在导航过程中的行为，对当前用于实现群机器人智能导航的一系列方法进行检查和评估。新的方法将根据结果创建。对于新方法的实施，将使用机器学习，试图将传统的人工智能与群体机器人相结合。从目前所做的文献综述来看，这似乎是一个尚未得到足够重视的任务（Brambilla et al., 2013）。因此，将产生的想法可以推动该领域向新的方向发展。群体智能算法将在简单的模拟环境中进行测试，以确保合作行为背后的基本思想得到正确实施。机器人平台的行为将使用机器人仿真软件进行模拟。在项目的后期阶段，软件将在真实的机器人平台上进行测试。在项目期间，将使用国家海洋学中心（NOC, 2017）提供的设备和设施，作为NEXUSS计划的一部分，该项目是该计划的一部分。目标：虽然群体机器人文献包含许多提出的导航算法，但在真实环境中进行测试的算法并不多（Brambilla等人，2013）（Tan和Zheng， 2013）。该项目旨在通过创建能够展示所使用算法功能的真实平台来填补这一空白。创建通用导航算法，用于其他群体机器人项目的开发。这将是在群体机器人领域实现这种规模的第一次尝试之一。挑战：平台复杂性/平台数量：为了使平台尽可能便宜和容易组装，它们的硬件复杂性必须受到限制。在这个项目中，这可以通过为蜂群使用更多的平台来补偿。另一方面，这会增加导航和通信任务的复杂性。因此，需要确定硬件复杂性和软件复杂性之间的适当比例。分布式处理/集中式处理：分布式处理意味着每个代理为自己做决定，这可能需要复杂的软件才能使群正常运行。另一方面，集中式处理表明其中一个代理是领导者，其他所有代理都服从它的命令。这也有它自己的挑战，比如领导者发生了一些事情，团队无法完成任务。确定系统在分布式/集中式处理频谱中的最佳位置将是这个项目的一大挑战。参考文献：Brambilla， M., Ferrante， E., Birattari， M.和Dorigo， M.（2013）。蜂群机器人：从蜂群工程的角度综述。群体智能，7(1),pp.1-41。谭勇，郑铮（2013）。群机器人技术的研究进展。国防科技，9(1),pp.18-39。NOC。(2017)。主页。检索自国家海洋学中心官方网站：https://noc.ac.uk/

This research project will aim to develop a swarm of autonomous marine vehicles that will be able to perform acoustic monitoring of the near-shore environment. This will include both the design of the hardware of the robotic systems and the implementation of the intelligent software that will allow the vehicles to perform the task in a cooperative manner.The robotic vehicles will make use of hydrophone sensors to passively monitor the marine environment, in order to detect and locate individual sources of noise below the surface of the sea (i.e. marine mammals, sub-marine vehicles, under-water turbines etc.).Initially, a range of current methods used to implement intelligent navigation of swarm robots will be examined and evaluated based the amount of information required to be exchanged, as well as the behaviour of the robots during navigation. New methods will be created based on the results.For the implementation of new methods, machine learning will be used, in an attempt to combine conventional artificial intelligence with swarm robotics. Based on the literature review done so far, it seems that this is a task that has not yet received enough attention (Brambilla et al., 2013). Therefore the ideas that will be produced can advance the field in new directions.The swarm intelligence algorithms will be tested using simple simulated environments, to ensure that the basic ideas behind the cooperative behaviour are implemented properly. The behaviour of the robotic platforms will be simulated using robot simulation software. In the later stages of the project, the software will be tested on real robotic platforms.During the project, there will be use of equipment and facilities provided by the National Oceanography Centre (NOC, 2017), as part of the NEXUSS program, that this project is part of.Objectives: Although the swarm robotics literature consists of many proposed navigation algorithms, not many have been tested in real environments (Brambilla et al., 2013) (Tan and Zheng, 2013). This project will aim to fill the gap by creating real platforms that will be able to demonstrate the capabilities of the algorithms used. The creation of a general navigation algorithm for use in the development of other swarm robotics projects. This will be one of the first attempts in achieving something of this scale in the field of swarm robotics.Challenges: Platform Complexity/ Platform Quantity: For the platforms to be as cheap and easily assembled as possible, their hardware complexity will need to be limited. For this project, this can be compensated by using a larger number of platforms for the swarm. On the other hand this can increase the complexity of the navigation and communication tasks. Therefore, a proper ratio between hardware complexity and software complexity needs to be identified. Distributed Processing/ Centralised Processing: Distributed processing suggests that each agent makes decisions for itself by itself, which can require complex software in order for the swarm to operate properly. On the other hand, centralised processing suggests that one of the agents is the leader and every other agent is subject to its commands. This can have its own challenges, like the case that something happens to the leader and the group is unable to finish the task. Identifying where the system should optimally lie in the distributed/centralised processing spectrum will be a big challenge of this project.References: Brambilla, M., Ferrante, E., Birattari, M. and Dorigo, M. (2013). Swarm robotics: a review from the swarm engineering perspective. Swarm Intelligence, 7(1), pp.1-41. Tan, Y. and Zheng, Z. (2013). Research Advance in Swarm Robotics. Defence Technology, 9(1), pp.18-39. NOC. (2017). Home Page. Retrieved from National Oceanography Centre Official Website: https://noc.ac.uk/