TTX3 Demonstrations and Posters

This research aims to develop an Ubiquitous Computing and Monitoring System (UCoMS) for oil/gas exploitation and management in the Gulf of Mexico. UCoMS address key research issues in wireless network systems, sensor technology, grid computing, and application software to enable the construction of a useful prototype for the discovery and management of energy resources in the Gulf of Mexico. The technical solutions resulting from this research will facilitate drilling and operational data logging and processing, on-platform information distribution and displaying, infrastructure monitoring/ intrusion detection, seismic processing and inversion, and management of complex surface facilities and pipelines. Decommissioned well platforms can be monitored and safeguarded using UCoMS with a potential of fostering new industries as well in the future. This research is a cooperative effort between research teams from The University of Louisiana at Lafayette, Louisiana State University, and Southern University and it is funded in part by the United States Department of Energy and by the Board of Regents of the State of Louisiana.

Building fires often spread to different areas very rapidly. It is also not always clear where the safest exits are, or which parts of the building may soon collapse. Therefore, we present SmokeNet, an active sensor network designed to provide evacuation assistance during a fire. It serves two main purposes: direct people inside a building to the safest exist, and give firefighters realtime updates of the fire. Unlike previous sensor network deployments, the purpose of SmokeNet is not simply data collection, but also in-network actuation.

In semiconductor manufacturing it is crucial to maintain a laminar air flow and a very clean environment. In fact, particles in the air can drop on wafers during processing. In order to keep the particles level very low, the air is circulated through filters that keep the air flow laminar and eliminate impurities. In a typical semiconductor plant there are thousands of filters that must be continuously monitored to insure air cleanliness. Differential pressure gauges are used to measure the pressure drop through the filters. Right now, facilities personnel walk through the floor and log the data manually on a spreadsheet for later analysis. This process is labor intensive and does not provide real time visibility of the filters status. We have developed a monitoring system based on MicaZ motes and Crossbow Moteworks™ network stack that collects the pressure data every second. All pressure sensors are battery powered and wirelessly connected to the base station. The sensors have local intelligence in order to immediately trigger an alarm in case the pressure measurement falls outside configurable bounds and otherwise cache a configurable number of measurements before sending them in one packet to the base station. We adopted a client-server architecture with a web browser user interface in order to access the data from virtually any device connected to the Internet. The objective was to have a real time system to monitor the filters, an analytical tool to schedule maintenance on demand and a higher degree of automation to reduce labor. The economical benefit would result from:

• less labor intensive process
• better filter utilization
• improved air quality control and consequently better manufacturing yield
• cost reduction due to improved filters utilization
• lower cost to relocate filters in case of clean-room re-layout due to the fact that the wireless pressure sensors can be easily moved to a different location

We present a novel radio interference based sensor localization method for wireless sensor networks. The technique relies on a pair of nodes emitting radio waves simultaneously at slightly different frequencies. The carrier frequency of the composite signal is between the two frequencies, but has a very low frequency envelope. Neighboring nodes can measure the energy of the envelope signal as the signal strength. The relative phase offset of this signal measured at two receivers is a function of the distances between the four nodes involved and the carrier frequency. By making multiple measurements in an at least 8-node network, it is possible to reconstruct the relative location of the nodes in 3D. Our prototype implementation on the MICA2 platform yields an average localization error as small as 3~cm and a range of up to 160~meters. In addition to this high precision and long range, the other main advantage of the Radio Interferometric Positioning System (RIPS) is the fact that it does not require any sensors other than the radio used for wireless communication.

Blackbook is a flash file system built using BlockStorage components in TinyOS. This file system currently runs on Telos motes, and was designed to allow systems built with Matchbox to be easily ported. Blackbook conserves memory in RAM and program flash, and uses a minimum amount of metadata for file entries in external flash. It is fault tolerant, wear-leveling, boots quickly, and also provides the ability to continuously log data to flash.

We developed a wireless sensor network for monitoring seismoacoustic activity at Volcán Reventador, Ecuador. Wireless sensor networks are a new technology and our group is among the first to apply them to monitoring volcanoes. The small size, low power, and wireless communication capabilities can greatly simplify deployments of large sensor arrays. The network consisted of 16 wireless sensor nodes, each outfitted with an 8 MHz CPU (TI MSP430) and a 4 GHz IEEE 4 radio (Chipcon CC2420) with data rates up to 80 Kbps. Each node acquired acoustic and seismic data at 24-bit resolution, with a microphone and either a single-axis geophone or triaxial short-period seismometer. Each node is powered by two D-cell batteries with a lifetime of about 1 week, and measures 18 x 10 x 8 cm. Nodes were distributed radially from the vent over a 3 km aperture. Control and data messages are relayed via radio to a base station node, with inter-node distances of up to 420 m. The base station transmits data using a FreeWave radio modem, via a repeater, to a laptop located 4 km from the deployment site. Each node samples continuous sensor data and a simple event-detection algorithm is used to trigger data collection. When a sensor detects an event, it relays a short message to the base station via radio. If several nodes report an event within a short time interval, the last 60 seconds of data is downloaded from each node in turn. One of the sensor nodes is programmed to transmit continuous data; due to limited radio bandwidth, it is not possible to collect continuous data from all nodes in the array. A GPS receiver and time synchronization protocol is used to establish a global timebase across all sensor nodes. The network operated for two weeks and triggered on over a hundred volcano-tectonic events, tremor episodes, and explosions. We are currently working to improve the system to scale up to much larger arrays and perform computations, such as eruption location and velocity inversion, on the sensor nodes themselves.

Delay/Fault Tolerant Mobile Sensor Network (DFT-MSN) distinguishes itself from conventional sensor networks by several unique characteristics, such as nodal mobility, sparse connectivity, delay tolerability, and fault tolerability, which make the data delivery schemes used in conventional sensor networks ineffective in DFT-MSN. In this work, we introduce a general communication protocol stack, which aims at providing an in-depth view of the protocol design for DFT-MSN. The protocol stack consists of four layers, i.e., physical layer, MAC layer, data delivery layer, and application layer. The lower layer provides services transparently to the upper layer, while the upper layer utilizes these services to fulfill specific functions. Our design of the stack focuses on the flexibility and modularity. Among these four protocol layers, the data transmission layer is the most unique and complex one. More specifically, a data delivery framework is introduced in the data transmission layer. This framework defines three modules, i.e., queue management module, message routing module, and mobility management module. The interfaces of each module are clearly clarified so that an existing module can be easily replaced by another compatible one. After describing the specific design considerations for each layer, we also exemplify the protocol stack with some popular protocols, in order to demonstrate the soundness of our design. With the layered architecture, the future researcher/developer can only focus on the design of his target layer/module, without much consideration of other layers/modules, which can significantly reduce the design and evaluation complexities.

After writing a wrapper for a closed source IEEE 802.15.4 stack for TinyOS, having to rewrite the existing TinyOS interfaces, we've decided to have a go at implementing our own open-source IEEE 802.15.4 stack written completely in NesC. The first challenge was to "decode" the IEEE 802.15.4 and ZigBee standards, in order to fully understand the details of the design needed. The design and requirements presented in the standards are in many ways in conflict with the traditional TinyOS way of thinking. It turns out, that the principles of buffer swapping and the simple synchronuous scheduling isn't always going to be adequate. This could be an indication, that more work still needs to be done in the TinyOS core system. Some of the key points of our design are:

• We provide generic interfaces that make use of abstract data types. This ensures easy adaptation of any existing stacks, as well as being user friendly.
• We make use of a buffer manager for allocation of both frame buffers and message primitives. At the same time we try to minimize overhead using lazy evaluation and early frame construction in order to avoid expensive memcpy operations.
• We aim to provide a stack small enough to be usable on the type of MCU's supported by TinyOS. From our experience, it seems that companies use the IEEE 802.15.4 enabled devices as a separate sub-system, using an entirely different platform to execute their main applications.

When evaluating the latest WSN protocols and applications,researchers must consider how to implement wireless conditions and device mobility. Often, they simulate--but research has shown that wireless simulation often gives unrealistic results because it does not fully capture the effects of radio signal propagation. Unfortunately, evaluation with real devices and real mobility becomes extremely difficult due to control and space issues. To lower these barriers to experimentation in the real world, we are continuing to develop wireless and mobility extensions to the Emulab network testbed. Our early deployment, available for public production use, provides six robots, each containing a small computer, 11 WiFi, and a Mica2 mote with a sensor board. Remote researchers can "drag and drop" robots interactively via a web interface as well as programmatically control their motion and network activity. We provide robot position data accurate to 1cm and path planning and obstacle avoidance algorithms that deal with low-level robot control. The robots move in an area containing 25 fixed sensor motes, which researchers can easily program and interact with via each mote's serial or ethernet connection. We have developed a low-cost power measurement device through which researchers can study mote power usage. In this demonstration, we'll show attendees how to interact with our testbed, and provide some sample experiments that they can tweak. We'll help attendees maneuver the robots in real-time via our web interface, and monitor (using streaming video and telemetry from individual robots) robot motion and WSN data from our experimental lab area in Utah.

Creating TinyOS applications from generic components often leads to applications containing dead code and constant variables. These unnecessarily consume SRAM and Flash memory. We address the problem with cXprop, a static-analysis-based post-processing tool for the C code produced by the nesC compiler. cXprop performs "conditional X propagation," a generalization of the well-known conditional constant propagation algorithm where X is an abstract value propagation domain supplied by the user. cXprop is interprocedural and achieves reasonable precision on pointer-rich codes. We use novel concurrency abstractions to analyze global state inside of TinyOS' restrictive concurrency model. The dataflow information produced by cXprop supports reduction in code size through interprocedural dead code elimination, and reduction in data size by finding global variables that do not use their full bitwidth. For example, state variables and boolean flags often use an entire byte to store a single bit. We have validated the code produced by cXprop through random testing and by dynamically checking its results against actual executions. cXprop reduces application code size on average by 10% for TinyOS applications and predicts savings of 50-100 bytes of SRAM. It is available as open source software.

New generations of sensor platforms have tracked technology trends in computation and communication components, but the storage components on these platforms have not kept pace. The high energy cost of storage has raised questions about the rationale of using in-network storage-based data management techniques for sensor networks. Our research aims at closing this gap. Identifying Energy-efficient Hardware: The low-power and low price associated with flash memory make it a natural choice for storage on sensors. We performed an exhaustive measurement of the energy consumptions of different flash memories and our results show parallel NAND flash to be up-to 100 times more energy-efficient. This changes conventional wisdom that considers storage cost to be equivalent to that of communication and in fact, makes it comparable to compute costs. Storage system Design: Most storage systems (YAFFS, ELF, etc.) and flash memories expose a file-system like interface. Though this is useful for mobile devices, it is hardly the interface of choice for the sensor domain since sensors typically do not operate on files. Additionally, the hardware limitations imposed by NAND flash technology motivates the design of a sensor-specific storage system. We have developed an object storage system based on NAND flash that exposes ‘storage objects’ like streams, indexes, etc. These objects are optimized for use on flash memory and provide an easy interface to stored sensor data. The system also supports basic transactions and data prioritization.

Wireless sensor monitoring application based on TinyOS technology. WSN nodes for this application are clone of Telos rev. B and MicaZ. Every WSN node transfers sensor reading values which is attatched on it. WSN embedded software is based on TinyOS Surge application. Server-side application displays tne network topology and sensor readings.

WLS(Wireless Light Switch), WSL access point are based on TinyOS and Telos rev. A. These are integrated with IBS light switch system, consist of LCU(Light Control Unit), master controller, light monitoring server. The WLS access point is a gateway between IEEE 802.15.4 and dedicated power line communication network. IBS light control system will be more flexible and easy-to-install at real deployment by this WLS system.

Luxoft Labs has developed eZeeNet, the network protocol stack software, based on ZigBee protocol. We have selected TinyOS for the eZeeNet development for a number of reasons. First of all, it is an open-source operating system developed with the wireless sensor network applications in mind. Its component based architecture enables rapid implementation while minimizing code size, as required by the low-memory constraints inherent in sensor networks. TinyOS' event-driven execution model enables precise power management while allowing the scheduling flexibility demanded by the unpredictable nature of wireless communication and physical world interfaces. However, while implementing MAC layer for the eZeeNet stack, we ran into some significant limitations of the current nesC compiler. Still committed to the TinyOS as a development platform, we decided to write a compiler version of our own that would address those shortcomings. Our goal was to create a comprehensive development platform for software development, debugging, testing, and simulation in the Wireless Sensor Network (WSN) field. At the heart of that platform is MeshC compiler, which is specifically geared towards wireless sensor applications. It introduces some revolutionary features such as algorithm optimization mechanism, advanced connection methods to wire new modules, and runtime wiring operators’ support. MeshC translator contains built-in algorithm analyzer that detects non-optimal algorithms in the MeshC code, and optimizes them. The module wirings implementation is not predefined. The developer can use any user-defined connection method to wire the modules into the application. That means that no source code modifications are needed.

We present TWIST, a scalable and flexible testbed architecture for indoor deployment of wireless sensor networks. The design of TWIST is based on an analysis of typical and desirable uses of sensor network testbeds. In addition to providing basic services like node configuration, network-wide programming, out-of-band extraction of debug data and gatewaying of application data, the architecture introduces several novel features, some of them exploiting the capabilities of the USB 0 standard. Firstly, TWIST supports different sensor node platforms. Secondly, it supports active power-control of the nodes. This enables easy transition between USB-powered and battery-powered experiments, dynamic selection of topologies as well as controlled injection of node-failures in the system. Thirdly, TWIST supports evaluation of both flat and hierarchical sensor networks. The "super nodes" that form the middle tier of the testbed infrastructure are dual-use devices that can also play a role as part of the sensor network application. The self-configuration capability, the use of standardized hardware and open-source software make the TWIST architecture scalable, affordable, and easily replicable. To demonstrate its practical value, we present our experiences with building and using a specific realization of the TWIST architecture that spans three floors of our office building and supports over one hundred sensor nodes.

There are many applications requiring entire set of data without loss. Structural health monitoring is one example. Reliable data collection (Straw) is implemented and evaluated. It works over a multi-hop network, is light-weight on a wireless node, provides high channel utilization and packet throughput, and collects data reliably. End-to-end communication is used guaranteeing end-to-end reliability. Most of complexity is in a receiver (powerful PC), and a sender is simple (wireless node). Parameters are optimized with network information, and rotating buffers enables an overlap of packet transmission and memory access. Packet throughput, bandwidth usage, and packet time are analyzed.

Building automation is considered to be one of three major markets for sensor networks. (The other two markets are home and industrial automation.) This demo will show a platform to help validate ubiquitous sensing concepts for building automation. The platform includes indoor environment sensors (relative humidity, light, barometric pressure, and temperature), a motion detector (passive infrared), and mobile activity sensors (two-axis accelerometer). The platform has analog and digital inputs to allow for rapid prototyping of third party devices. In addition to being able to sense, the sensor platform can actuate external devices through two relays. The platform’s modular hardware architecture allows it to use any one of Crossbow’s MICA Motes (315, 433, 915, 2400 MHz). The demonstration will use the MoteView user interface to log and display sensor values.

Several empirical studies in recent years have demonstrated the need for enhanced simulation models that take into account realistic wireless link conditions. It is of particular importance to develop simple parameterized models that can generate probabilistic packet losses based on input parameters for radio settings, the environment, and hardware variation. Our group has been developing such models for wireless link quality both without and with interference. We are working to incorporate these models into existing simulation environments such as tossim and emstar.

With increasing diversity arising in sensor networking applications and hardware, the availability of an extensible and intuitive set of common design tools has the potential to fulfill a wide variety of application and system needs. Sensor network deployment scale is steadily increasing, augmented by larger, more capable sensor systems, in turn creating higher-level processing ability. This creates the opportunity for the realization of innovative applications with very specific needs and features. Among these kinds of applications, some expected key needs include: . Low-level access and high-level abstractions to a wide variety of sensing hardware & platforms (i.e. various motes, video sensors, industrial control systems), . Flexible and simple communications for data and control (RPC/IPC for control and a basic raw transfer protocol for data), and . Simple/easy development, deployment, management, and re-tasking in mobile and ad-hoc settings. Cascades, written in the scripted (platform-independent) language Python, aims to provide a solution to these needs using a rich toolset for communication, data processing, and interfacing with various devices. Our presentation conveys the project's current progress through a set of basic ideas and their applications.

A 30-mote network was deployed in an experimental forest of National Taiwan University to collect data for microclimate analysis. While the research is still ongoing after data has been collected and analyzed over the fall and winter months, the initial results have shown that wireless sensor network technology will become one of the most viable frameworks for large scale field research from all aspects.

A wireless, high-resolution galvanostat is fabricated using "smart dust" technology for under $100 dollars using free software and general use hardware. We demonstrate a design that uses either PWM or DAC to set a current with up to 100-picoampere resolution for any value between -15 and 15 mA. An ADC measures the potential of a battery during charge cycles using a short (~400 µs) interrupt. The device performs well when tested against calibrated values for a variety of capacitors and resistors.

The motes in current wireless sensor networks employ custom hardware with little peripheral device support. The main issues with that are: (1) The lack of input/output devices such as keyboard and display makes it hard to develop and debug (2) There is no standard interface between the microcontroller boards and the sensor add-on boards so the boards from different companies are not interchangeable. To address those issues we are trying to use off-the-shelf handheld PDAs to construct a sensor network. A PDA is basically a complete computer system so it makes development very effective and yet when all the peripherals are stripped, it becomes a mote. In fact, the latest version of motes from the Intel Lab uses the same CPU as that used in Dell and HP PDAs. We will also show some sensor boards (humidity sensors, magnetometers and accelerometers) using standard CompactFlash Plus (CF+) interface.

We demonstrate a system for location estimation using Bayesian hierarchical models. Our approach achieves accuracy similar to other models and algorithms (2m-3m). In addition, we demonstate a novel sampling technique that can solve these Bayesian networks in real-time. With a sufficent number of clients, we also show our approach can eliminate the requirement for training data, thereby introducing a fully adaptive zero profiling system for location estimation.

A challenge for many wireless sensor networks is to remain operational for long periods of time on a very limited power supply. While many power management protocols have been proposed, a solution does not yet exist that allows them to be seamlessly integrated into the existing systems. In this paper we study the architectural support required to resolve this issue. We propose a framework that separates sleep scheduling from the basic MAC layer functionality and provide a set of unified interfaces between them. This framework enables different sleep scheduling policies to be easily implemented on top of multiple MAC layers. Such a flexibility allows applications to choose the best sleep scheduling policy based on their own particular needs. We demonstrate the practicality of our approach by implementing this framework on top of both the mica2 and telosb radio stacks in TinyOS Our micro-benchmark results show that at the cost of a slight increase in code size, our framework significantly eases the development of new radio power management protocols across multiple WSN platforms.

Moteiv will demonstrate Tmote Invent, our new, low cost wireless sensing system. Tmote Invent includes Moteiv's popular Tmote Sky module for computation and communication, light, temperature, sound, and acceleration sensors, a speaker, and a rechargeable battery packaged in an innovative case design. Demonstrations of the many potential applications that can be built with Tmote Invent will be shown; these include streaming audio, industrial vibration monitoring, and ultra low power large scale networking with IEEE 802.15.4. Under the hood, Tmote Invent runs Moteiv-branded TinyOS, a highly stress-tested distribution of the popular open source operating system that combines the most cutting-edge ideas from the open source community with Moteiv-certified highly reliable, highly predictable, low power software.

Typical sensornet deployments are focused on many-to-one and one-to-many routing schemes. Point-to-point routing has received less attention in sensornet research; however, recent applications have become more sophisticated, and therefore they require the ability to communicate with, and control, specific nodes (e.g. purser-evader game). Thus, it's crucial to enable point-to-point routing. Authors in [1] designed and implemented an algorithm called Beacon Vector Routing (BVR). BVR has a set of special nodes in the network called beacons. BVR creates a dynamic coordinate system based on connectivity to these beacons that allows nodes to send messages to a coordinate thus enabling point-to-point routing. Crucial to BVR is the destination node's coordinates. An application using the send/receive interface of BVR should refer to nodes by an identifier, rather than coordinates (coordinates are dynamic due to node failure and link availability). To accomplish this task, a location service should exist between the application and BVR to translate node identifier to node coordinates. This service is the focus of our work. We have devised a scheme which uses beacons as storage places for these identifier-to-coordinate mappings. Each node periodically publishes its coordinates to the appropriate beacon using a global hash function. After the location-service registration phase, nodes may route data to each other by hashing the target node's identifier and querying the beacon whose hash value most closely matches the hash output of the target node. We have implemented an initial version of the scheme and tested it in simulation and look to continue our investigation in a test-bed environment. [1] Rodrigo Fonseca, Sylvia Ratnasamy, Jerry Zhao, Cheng Tien Ee, David Culler, Scott Shenker, and Ion Stoica. Beacon vector routing: Scalable point-to-point in wireless sensornets. In Proceedings of the 2nd Symposium on Networked Systems Design and Implementation (NSDI '05), pages 329-342, 2005.

The Re-Mote Testbed Framework is an alternative to MoteLab and Mirage. The main focus points of the system design are: Modular design, abstraction levels, and responsiveness. The components of the framework fall into three categories: Client software, server software and host/gateway software. The client is a Java based GUI featuring mote console windows useful for debugging applications. The server software consists of a mote server daemon written in C++, a relational database based on MySQL, and a mote information server based on Apache Tomcat. The host software is application dependent. In the actual deployment there is a host daemon keeping track of connected motes, along with a customized USB to serial driver and a flash programming utility. A proof-of-concept Re-Mote Testbed Framework with about 40 Freescale DIG-528 boards was deployed at Department of Computer Science, University of Copenhagen in fall 2005. The students at Philippe Bonnet's Sensor Networks course successfully used the testbed for their experiments. A full scale deployment with about 120 boards is planned for 2006.

Sensor Networks provide a cheap, unobtrusive, and easy to deploy method for gathering large amounts of data from an environment. However, this data is often significantly noisy. One way to improve the confidence in the data is to build a model of the correlations among nearby sensor readings and do inference on this model to derive data with less error. Previous methods of inference on sensor networks require large amounts of communication to move the data to a central base station or are intractable for many classes of models. We propose using Tree-Reweighted Loopy Belief Propagation, a less communication intensive inference algorithm that is well suited to the distributed nature of Sensor Networks. We discuss implementation issues such as lowering communication costs by modifying scheduling of message transmissions. We also investigate various modes of failure such as non-functional motes and asymmetric communication. We present simulated results to compare the scheduling schemes, as well as evaluate the effects of network failures. We also implemented a version of belief propagation that runs over Sensor Networks and verified the results are the same as the simulated version. Ultimately, we show that Tree-Reweighted Loopy Belief Propagation is well suited for sensor fusion on Sensor Networks.

Residential energy consumption accounts for nearly 20% of the total national consumption and 50% of the critical peak consumption in California. With blackout-related economic losses costing billions of dollars, residential energy consumption is at the forefront of energy management research. This research focuses on utilizing wireless sensor networks as an enabling technology for novel energy management applications. Such applications require extreme deployment lengths in excess of a decade, in which battery replacement is not optional. Design requires an integrated approach to balance performance of the application, network, and node. Specifically, comfort-based residential HVAC control strategies that leverage distributed sensing, characterizing 2.4GHz wireless sensor network communication performance in residential settings, evaluating energy scavenging opportunities in residential environments, and designing application-specific energy scavenging power systems.

We present a robust embedded software suite built on the open-source TinyOS 2.x operating system, running on the Moteiv Tmote Sky, Crossbow MicaZ, and Intel Mote2 platforms. It provides a rich Web Services interface for all the capabilities of the sensor network. Arched Rock software bridges the gap between the physical world and the Internet.