ITR: Scalable Location-Aware Monitoring (SLAM) Systems

这份提案描述了SLAM，这是一种可扩展的网络体系结构，将数百万真实世界的传感器与执行器和分布式软件应用程序集成在一起。SLAM将实现各种新颖的监测和控制应用，包括快速灾难响应、可扩展的犯罪检测和预防、设施维护、资产监控和导航。SLAM解决了三个问题：1.充分利用传感器的数据流需要了解上下文信息，特别是位置和时间。对数以百万计的资产和设施进行精细监控需要在环境中实际部署传感器，这是一项密集而繁琐的人工任务。分布式软件应用程序使用部署的传感器/执行器需要网络基础设施。SLAM体系结构有三个主要组件来解决这些问题：1.Cricket，一个无处不在的精确定位基础设施。目前没有一种位置感知技术在任何地方、任何时间、任何地方都有效。Cricket是一种解决这一问题的新型多传感器定位架构，它结合了室内和建筑物周边的射频和超声波，以及室外的GPS。Cricket结合了自我配置算法和节能协议，以实现可伸缩性和寿命。2.激活环境和有效的激活方法。SLAM要求用传感器和执行器激活主体环境。如果不特别注意，激活过程可能会由于环境的复杂性而变得无法管理。因此，SLAM提供了基于虚拟位置的标记，通常用于固定对象。人工安装人员通过使用配备Cricket的手持设备指向物理区域或对象来将虚拟标签附加到物理区域或对象，从而触发永久存储中唯一标识符与被标记实体的位置和其他属性的关联。这简化了环境激活。可扩展的网络基础设施将传感器信息和事件连接到软件处理程序。网络由固定的和移动的传感器代理组成，与它们监视的对象和事件物理上位于同一位置，以集成位置、身份和时间信息以形成事件流。传感器及其代理使用特定于传感器的低能量通信协议进行通信。应用程序被编写为分布在网络上的事件处理程序。SLAM支持跨代理和计算服务器动态分发处理程序、将事件路由到处理程序以及执行查询处理操作。建议的SLAM体系结构引入了三个创新的想法：无处不在的、高能效的位置基础设施(借鉴了基于信标的定位系统、计算几何和无线网络的想法)；用于环境激活和资产管理的虚拟区域和对象标记(借鉴了几何建模和数据库管理系统的想法)；以及基于分布式代理的事件和响应处理(借鉴了网络和数据库系统的想法)。从现有环境(建筑物、校园或城镇)开始，实施SLAM的操作模型如下。首先，位置基础设施被激活。将位置信标放置在环境中，并构建环境的数字表示，从而能够在环境中的任何位置进行位置推断。第二，环境被激活。传感器以及虚拟和物理标签被贴在环境(和环境表示)内的感兴趣的对象上。第三，SLAM网络被激活，将原始传感器数据流连接到传感器代理。代理使用位置和时间信息注释传感器数据流，并通过事件处理网络将其转发到适当的处理程序。处理器产生更多的事件，以及要转发给执行器或人类的动作和通知。作为一个具有挑战性的测试案例，我们计划在一个拥有数百万个有趣实体的大型大学校园中部署SLAM。其中包括办公室、机房、物理厂房和实验室中的许多传感器，用于监控电力、温度、湿度和压力；烟雾和火灾探测器；防盗报警器和物理入侵探测系统；运动探测器；泄漏、洪水、化学品和危险材料的监视器；大型防盗和犯罪设备以及导航辅助设备。目标是监控该大学的实物资产，并提高数以百计建筑的数千个办公室、实验室和公共空间内及其周围活动人员的人身安全。目标SLAM系统最初将专注于麻省理工学院与各种感兴趣的合作伙伴的三项功能：高效的设施监控和维护(与麻省理工学院实体工厂合作)；可扩展的资产监控以进行清点、预防和侦测(与麻省理工学院财产办公室、麻省理工学院校园警察和麻省理工学院图书馆合作)；导航辅助，包括个人寻路和无处不在的主动标志(与麻省理工学院时间表办公室和麻省理工学院安全办公室合作)。

This proposal describes SLAM, a scalable network architecture integrating millions of real-world sensors with actuators and distributed software applications. SLAM will enable a broad variety of novel monitoring and control applications including rapid disaster response, scalable crime detection and prevention, facilities maintenance, asset monitoring, and navigation. SLAM solves three problems:1. Full exploitation of a sensor's data stream requires knowledge of contextual information, particularly location and time.2. Fine-grained monitoring of millions of assets and facilities requires the physical deployment of sensors in the environment an intensive and cumbersome manual task.3. Use of deployed sensors/actuators by distributed software applications requires network infrastructure.The SLAM architecture has three main components that address these issues:1. Cricket, a ubiquitous and precise location infrastructure. No current location-sensing technology works everywhere in all places and at all times. Cricket is a novel multi-sensor location architecture to solve this problem, using a combination of RF and ultrasound indoors and at building perimeters, and GPS outdoors. Cricket incorporates self-configuration algorithms and energy-efficient protocols for scalability and longevity. 2. An activated environment and efficient activation method. SLAM requires that the subject environment be activated with sensors and actuators. Without special attention, the activation process could become unmanageable due to the complexity of the environment. Therefore SLAM provides virtual location-based tagging, typically for immobile objects. The human installer affixes virtual tags to physical regions or objects by pointing at them with a Cricket-equipped handheld device, triggering an association of a unique identifier and the tagged entity's location and other attributes in a persistent store. This eases environment activation.3. A scalable network infrastructure connects sensor information and events to software handlers. The network consists of fixed and mobile sensor proxies, physically co-located with the objects and events they monitor, to integrate location, identity, and temporal information to form an event stream. Sensors and their proxies communicate using sensor-specific low-energy communication protocols. Applications are written as event handlers distributed across the network. SLAM provides support for dynamically distributing handlers across proxies and compute servers, routing events to handlers, and performing query processing operations. The proposed SLAM architecture introduces three innovative ideas: ubiquitous, energy-efficient location infrastructure (drawing on ideas from beacon-based location systems, computational geometry, and wireless networking); virtual region and object tagging for environment activation and asset management (drawing on ideas from geometric modeling and database management systems); and distributed proxy-based event and response processing (drawing on ideas from networking and database systems).Starting with an existing environment (a building, campus, or town), the operational model to put SLAM in place is as follows. First, the location infrastructure is activated. Location beacons are placed in the environment, and a digital representation of the environment is constructed, enabling location inference anywhere within the environment. Second, the environment is activated. Sensors and virtual and physical tags are affixed to objects of interest within the environment (and environment representation). Third, the SLAM network is activated, connecting raw sensordata streams to sensor proxies. The proxies annotate sensor data streams with location and temporal information, and forward them to appropriate handlers via the event-processing network. Handlers produce further events, as well as actions and notifications to be forwarded to actuators or humans.As a challenging test case, we plan to deploy SLAM on a large university campus with millions of interesting entities. These include many sensors in offices, machine rooms, physical plant, and laboratories to monitor power, temperature, humidity, and pressure; smoke and fire detectors; burglar alarms and physical intrusion detection systems; motion detectors; monitors of leaks, floods, chemicals, and hazardous materials; large-scale theft- and crime-prevention apparatus, and navigation aids. The goal is to monitor the university's physical assets and improve the personal safety of over ten thousand individuals moving in and around thousands of offices, labs, and common spaces in hundreds of buildings.The target SLAM system will focus initially on three capabilities at MIT with a variety of interested partners: efficient facilities monitoring and maintenance (with MIT Physical Plant); scalable asset monitoring for inventory, crime prevention and detection (with the MIT Property Office, MIT Campus Police, and MIT Libraries); and navigation assistance, including both personal way-finding and pervasive active signage (with the MIT Schedules Office and the MIT Safety Office).