CENS, a NSF Science & Technology Center, is developing Embedded Networked Sensing Systems and applying this revolutionary technology to critical scientific and social applications. Like the Internet, these large-scale, distributed, systems, composed of smart sensors and actuators embedded in the physical world, will eventually infuse the entire world, but at a physical level instead of virtual. An interdisciplinary and multi-institutional venture, CENS involves hundreds of faculty, engineers, graduate student researchers, and undergraduate students from multiple disciplines at the partner institutions of University of California at Los Angeles (UCLA), University of Southern California (USC), University of California Riverside (UCR), California Institute of Technology (Caltech), University of California at Merced (UCM), and California State University at Los Angeles (CSULA).
This project investigates the development of undersea sensor networks. We highlight potential applications to off-shore oilfields for seismic monitoring, equipment monitoring, and underwater robotics. We identify research directions in short-range acoustic communications, MAC, time synchronization, and localization protocols for high-latency acoustic networks, long-duration network sleeping, and application-level data scheduling. We describe our preliminary design on short-range acoustic communication hardware, and summarize results of high-latency time synchronization.
Tiered systems comprised of low-power, resource constrained sensor devices (motes) and higher powered microservers are quickly becoming a reality. A typical deployment scenario would consist of several microserver-class devices, each of which would be responsible for a set of motes. Mote clustering is a set of tools and services that enables interoperation between motes and microservers, thereby allowing applications to seamlessly use the available hardware infrastructure in order to achieve their task. In particular, the goal is to leverage the computational power of the microservers, as well as the higher quality state view (in terms of both temporal and spatial resolution). One of the driving applications behind mote clustering is flexible triggering of imagers. This will be comprised of a collection of microservers with cameras attached to them, as well as a deployment of motes. The motes would sense the environment and trigger the cameras when something of interest is detected. Using mote herding, the application can associate mote locations and camera fields of view as well as locate the appropriate camera for each task.
We propose a vision-based SLAM algorithm incorporating feature descriptors derived from multiple views of a scene, incorporating illumination and viewpoint variations. These descriptors are extracted from video and then applied to the challenging task of wide baseline matching across significant viewpoint changes. The system incorporates a single camera on a mobile robot in an extended Kalman filter framework to develop a 3D map of the environment and determine egomotion. At the same time, the feature descriptors are generated from the video sequence, which can be used to localize the robot when it returns to a mapped location. The kidnapped robot problem is addressed by matching descriptors without any estimate of position, then determining the epipolar geometry with respect to a known position in the map.
Advances in technology and infrastructure have positioned mobile phones as a convenient platform for real-time assessment of an individuals health and behavior, while offering unprecedented accessibility and affordability to both the producers and the consumers of the data. In this paper we address several of the key challenges that arise in leveraging smartphones for health: designing the complex set of building blocks required for an end-to-end system, motivating participants to sustain engagement in long-lived data collection, and interpreting both the data and the quality of the data collected.
We present AndWellness, a mobile to web platform that records, analyzes, and visualizes data from both prompted experience samples entered by the user, as well as continuous streams of data passively collected from sensors onboard the mobile device. In order to address the system design and participation motivation challenges, we have incorporated feedback from hundreds of behavioral and technology researchers, focus group participants, and end-users of the system in an iterative design process. AndWellness additionally includes rich system and user analytics to instrument the act of participation itself and ultimately to contextualize and better understand the factors affecting the quality of collected data over time. We evaluate the usability and feasibility of AndWellness using data from 3 studies with a variety of populations including young moms and recent breast cancer survivors. More than 85% of the diverse set of participants who responded to exit surveys claim they would use AndWellness for further personal behavior discovery.
Wireless sensor networks have attracted attention from a diverse set of researchers, due to the unique combination of distributed, resource and data processing constraints. However, until now, the lack of real sensor network deployments have resulted in ad-hoc assumptions on a wide range of issues including topology characteristics and data distribution. As deployments of sensor networks become more widespread [1, 2], many of these assumptions need to be revisited. This paper deals with the fundamental issue of spatio-temporal irregularity in sensor networks We make the case for the existence of such irregular spatio-temporal sampling, and show that it impacts many performance issues in sensor networks. For instance, data aggregation schemes provide inaccurate results, compression efficiency is dramatically reduced, data storage skews storage load among nodes and incurs significantly greater routing overhead. To mitigate the impact of irregularity, we outline a spectrum of solutions. For data aggregation and compression, we propose the use of spatial interpolation of data (first suggested by Ganeriwal et al in ) and temporal signal segmentation followed by alignment. To reduce the cost of data-centric storage and routing, we propose the use of virtualization, and boundary detection.
In this paper, we present a platform for collaborative acoustic signal processing, and demonstrate its use with an example application. Our platform is built upon the Stargate Linux-based microserver, and supports synchronized multi-channel acoustic data acquisition. We implement a dataflow-like staged event-driven programming model within the Emstar software framework that simplifies the development of collaborative processing applications. Unlike previous dataflow systems that emphasize real-time constraints, our framework emphasizes collaborative processing across nodes in a distributed system connected by an energy-conserving wireless network with non-deterministic message latency. In our model, an application is constructed by wiring together multiple stages, where each stage is implemented by an EmStar module. The modular approach simplifies development by isolating errors to specific stages, and enables run-time systemreconfigurability by allowing users to swap out implementations of individual stages, and to reconfigure the dataflow at run time.
A model of optimally precise and globally consistent clock synchronization, using the model provided by Reference-Broadcast Synchronization.
We propose an entropy-based sensor selection heuristic for localization. Given 1) a prior probability distribution of the target location, and 2) the locations and the sensing models of a set of candidate sensors for selection, the heuristic selects an informative sensor such that the fusion of the selected sensor observation with the prior target location distribution would yield on average the greatest or nearly the greatest reduction in the entropy of the target location distribution. The heuristic greedily selects one sensor in each step without retrieving any actual sensor observations. The heuristic is also computationally much simpler than the mutual-information-based approaches. The effectiveness of the heuristic is evaluated using localization simulations in which Gaussian sensing models are assumed for simplicity. The heuristic is more effective when the optimal candidate sensor is more informative.
A Monolithically Fabricated Combinatorial Mixer for Microchip-Based High-Throughput Cell Culturing Assays
We present an integrated method to fabricate 3-D microfluidic networks and fabricated the first on-chip cell culture device with an integrated combinatorial mixer. The combinatorial mixer is designed for screening the combinatorial effects of different compounds on cells. The monolithic fabrication method with parylene C as the basic structural material allows us to avoid wafer bonding and achieves precise alignment between microfluidic channels. As a proof-of-concept, we fabricated a device with a three-input combinatorial mixer and demonstrated that the mixer can produce all the possible combinations. Also, we demonstrated the ability to culture cells on-chip and performed a simple cell assay on-chip using trypan blue to stain dead cells.
We present the design, implementation, and evaluation of the Acoustic Embedded Networked Sensing Box (ENSBox), a platform for prototyping rapid-deployable distributed acoustic sensing systems, particularly distributed source localization. Each ENSBox integrates an ARM processor running Linux and supports key facilities required for source localization: a sensor array, wireless network services, time synchronization, and precise self-calibration of array position and orientation. The ENSBox’s integrated, high precision self-calibration facility sets it apart from other platforms. This self-calibration is precise enough to support acoustic source localization applications in complex, realistic environments: e.g., 5 cm average 2D position error and 1.5 degree average orientation error over a partially obstructed 80x50 m outdoor area. Further, our integration of array orientation into the position estimation algorithm is a novel extension of traditional multilateration techniques. We present the result of several different test deployments, measuring the performance of the system in urban settings, as well as forested, hilly environments with obstructing foliage and 20–30 m distances between neighboring nodes.