Data analytics approach to design verification of distributed systems and sensor networks

传感器网络处于先进分析系统的前沿。通过在许多应用领域部署机器学习支持的传感器，传感器网络变得越来越智能，包括战术系统，智能交通，医疗保健，环境监测，石油钻探和智能配电。基于智能传感器网络的应用系统（即基础设施、硬件和软件）通常是开放式的，并在几次迭代中逐步开发。开放意味着系统必须科普多余/冲突的需求。设计这样的系统的理论基础已经在分布式系统（DS）的研究，其中功能和/或控制是分布式的。在软件工程和人工智能中，如分布式软件系统（DSS）和多智能体系统（MAS），详细讨论了DS的相关理论及其实际实现。在DS中，组件（例如传感器，代理）交互的方式通常由场景（例如序列图）描述。在大规模系统中，可能存在数千个这样的场景。在设计和开发的多个迭代中维护场景之间的一致性是一项复杂且昂贵的任务。例如，在商用无人飞行器（UAV）机群中，每个UAV中存在若干传感器，并且异构UAV机群可以具有不同的运动场景和任务分配。无人机之间的充分通信使协调和动态任务分配成为可能。无人机被组装成一个舰队，逐渐增加。由于缺乏集中控制和场景的多样性，整个系统可能会表现出非预期/意外的行为，通常称为组件级（例如，在每个UAV内）的“紧急行为”（EB）和系统级（例如，机队中的UAV）的“隐含场景”（IS）。EB/IS可能会对用户、环境和业务造成代价高昂和/或不可逆转的损害。 我们在这项研究中的目标是管理（即建模，分析，检测和解决）不必要的行为（即EB和IS），通过识别设计缺陷，可能会导致不必要的行为，尽早在系统开发过程中。 本研究的独特之处在于：（1）将组件之间的交互和组件内部的状态结合起来建模，保存了组件的状态，并保留了组件之间的交互信息;（2）利用交互信息来检测潜在的问题;（3）能够根据识别出的交互来研究组件之间是否存在新的行为;（4）能够根据检测到的问题的原因提出解决方案;（5）能够调查传感器网络中常见的同类组件的相互作用。 典型的应用领域-对加拿大极其重要-是智能交通，能源，健康和机器人等。

Sensor networks are at the front line of advanced analytic systems. Sensor networks are getting smarter through deployment of machine learning empowered sensors in many application areas including tactical systems, intelligent transportation, health care, environmental monitoring, oil drilling and smart power distribution. Application systems (i.e. infrastructure, hardware and software) based on smart sensor networks are usually open ended and developed incrementally in several iterations. Being open implies that the system must cope with superfluous/conflicting requirements. Theoretical basis of designing such systems has been studied in the distributed systems (DS) research in which functionality and/or control are distributed. DS related theories and their practical implementations are discussed in detail in software engineering and artificial intelligence, e.g. distributed software systems (DSS) and multi-agent systems (MAS). In DS the way components (e.g. sensors, agents) interact is usually described by scenarios (e.g. sequence diagrams). In a large scale system, thousands of such scenarios may exist. Maintaining consistency among scenarios in multiple iteration of design and development is a complex and expensive task. For example, in a commercial unmanned aerial vehicle (UAV) fleet, there are several sensors in each UAV and a fleet of heterogeneous UAVs may have different motion scenarios and task allocations. The full communication between the UAVs enables coordination and dynamic task allocation. The UAVs are assembled as a fleet, incrementally. Due to lack of centralized control and multiplicity of scenarios, the overall system may exhibit unintended/unexpected behavior, commonly known as “emergent behavior” (EB) at the component level (e.g. within each UAV) and “implied scenario” (IS) at the system level (e.g. UAVs in a fleet). EB/IS may lead to costly and/or irreversible damage to the users, environment, and the business. Our goal in this research is to manage (i.e. model, analyze, detect and resolve) unwanted behavior (i.e. EB and IS) by identifying design flaws that may lead to unwanted behavior as early as possible in the system development process. Unique points in this research are: (1) Modeling both interactions of components and their internal states together that saves the components' states and preserves the interaction information among the components; (2) Using interaction information to detect potential problems; (3) Ability to investigate whether new behavior can exist between components based on identified interactions; (4) Ability to suggest solutions based on the cause of the detected problem; (5) Ability to investigate interactions of the same-type components, which are common in sensor networks. Typical applications areas - extremely important to Canada - are intelligent transportation, energy, health and robotics, among the others.