Localization is an important part of sensor network deployment, as it gives physical context to data which nodes sense. During my undergraduate and early graduate career, I developed a simple visualiser and system to range and localize nodes using acoustic ranging and multi-lateration algorithms. This system helped me gather data for my undergraduate work, and provided an interactive demonstration to visitors. The mote software is developed in TinyOS/nesC and the visualisation software is developed in Java. I used the Vanderbilt University ranging library as a base for my work, which required me to develop application level software, as well as low-level data gathering components. I made the demonstration visual and interactive by allowing node localization to be movement prompted. In order to initialise ranging (and thus localization), someone could shake it, place it down and then leave it to range with its neighbours, estimating position. This meant we could investigate iterative localization starting with only 3 sensors with known reference positions.
Related Demos
A video of the localization is held by Cogent, which accompanied my 2005 undergraduate thesis. I also created a simple messaging and LED counting application to show prospective students as part of a Cogent Open Day presentation at Coventry University. Nodes would pass around a token, whoever had the token would make an acoustic signal and then pass it on to a neighbour. This was intended to demonstrate some simple capabilities of the nodes and sensors.
I attempted to generalise my localization work to 3D, using the same multilateration techniques and Mica2 hardware. This revealed the limitations of using Mica2 motes, and their sensor board for real-life localization applications. The ranging and localization algorithms were highly affected by reverberations in the indoor experimentation environment, resulting in poor iterative localization performance.
Related Publications
Allen, M., Gaura, E., Newman, R., Mount, S., "Experimental Localization with MICA2 Motes", In Proceedings of NSTI Nanotech 2006, pp 435-440
Allen, M., Gaura, E., Newman, R., Mount, S., "Experimental Ranging with MICA2 Motes", In Proceedings of the 2nd ACM Workshop on Real-World Wireless Sensor Networks , RealWSN ‘06
I collaborated with Tom Schoellhammer at CENS, UCLA during the summer of 2006. I looked at tools that could aid deployment localization (using GPS) as well as real-time, interactions with nodes to monitor:
I implemented a GTK-based user interface on the Nokia 770 (with large buttons for touch screen operation), and ported EmStar to the platform (so it could host a mica2 mote as its platform). This became the 'deployment PDA', which allowed a user to mark and store GPS positions of nodes (gathered from a bluetooth GPS device paired with the 770), and query nodes within one hop for messages left by others (MNotes), as well as currently running ESS queries (GetQuery). Short term link estimation (to aid placement) was facilitated using Marco Polo, which sent ping-like messages - proxying messages from the nearest node to its one-hop neighbours.
Our work culminated in a sensing deployment at James Reserve, on the AMARSS transect. We augmented the transect with 11 mica2 nodes, gathering soil moisture and temperature (at three depths - 2cm, 8cm and 12cm). Several CENS wiki pages document my development of these components as well as the James Reserve deployment.
I collaborate on the ongoing acoustic sensing work at UCLA, an inter-departmental effort bringing biologists, electrical engineers and computer scientists together to address acoustic detection, classification and localization projects. We work with the Acoustic Embedded Networked Sensing Box (ENSBox), developed by Lewis Girod and Travis Collier, a linux based embedded platform for acoustic sensing. Software for the platform is traditionally written using the EmStar, framework.
I developed an on-line demonstration of the work presented in IPSN '07, where acoustic ensboxes were used to localize a whistle in an on-line fashion. I used EmStar, and integrated the components described in the paper such that they could work on-line, developing or porting code from Matlab where required. A video of the system running is on youtube (here). This demo was shown at IPSN 2007 and the 2007 CENS site visit. Python visualisation was implemented by Travis Collier, and I implemented web visualisation.
This work led me to working with the Wavescope project, and VoxNet in particular.
Related Publications
Allen, M.P., Graham, E., Ahmadian, S., Ko, T. Yuen, E., Girod, L., Hamilton, M., Estrin, D., "Interactive Environmental Sensing: Signal and Image Processing Challenges", IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP 2008)
I am an active user and contributor to EmStar at CENS, UCLA. This is mainly due to my involvement with the acoustic sensing project at CENS. EmStar is a software development framework for wireless network software, typically with higher capability 'microserver' class devices.
The acoustic laptop is a port of the acoustic ensbox software (written in EmStar), which allows the user to run a laptop as if it were an acoustic ensbox. This platform is not intended to replace the Acoustic ENSBox, but to provide a development stepping stone between simulation, emulation and real-life testing. This platform aids development, and has been useful for other users, namely Prof. Yao's group at EE, UCLA (to perform data collection for 3D direction of arrival estimation algorithms) and the HPWREN group at UCSD, (led by Hans-Werner Braun) who used an x86 based Soekris platform to make a standalone acoustic array for monitoring wolf calls (pictures here and here).
Related Publications
Allen, M. Girod, L., Estrin, D. "Acoustic Laptops as a research enabler", In Proceedings of the Fourth Workshop on embedded Networked Sensors, EmNets 2007.
Related Links: Acoustic Laptop Installation
VoxNet is a platform for distributed acoustic sensing, encompassing both hardware and software. I worked on developing and integrating various aspects of the system together (spill to disk, multi-hop routing, wavescope shell, dissemination), as well as some application level components, notably some stream visualiser software in Java, using the JFreeChart libraries. This software allows the user to visualise data being transferred over wavescope network streams. Because it is java-based, it can potentially be used on PDA type devices, allowing the user to un-tether from the main control point in VoxNet, and still get feedback as to the system/application's current state.
We deployed our VoxNet platform on an exploratory basis at the Rocky Mountain Biological Laboratory, Colorado, in summer 2007.
Related Publications
Allen, M., Girod, L., Newton, R., Madden, S., Blumstein, D., Estrin, D., "VoxNet: An Interactive, Rapidly-Deployable Acoustic Monitoring Platform", International Conference on Information Processing in Sensor Networks (IPSN 2008)
Demos
We demonstrated a simple VoxNet application at the 2008 CENS site visit at UCLA, and the 2008 WiSig demo Showcase at NPL. The demo detected hand claps and estimated direction of arrival, as well as streaming audio data over the network. Visualisation was performed using Baudline and GNUPlot.
I collaborate on the Wavescope project at MIT. This introduction is taken from the Wavescope website:
"Wavescope is a system for developing distributed, high-rate applications that need to process streams of data from various sources (e.g., sensors) using a combination of signal processing and database (event stream processing) operations. The execution environment for these applications ranges from embedded sensor nodes to multicore/multiprocessor servers." Wavescope programs are written in Wavescript, a high-level stream processing language in which a directed graph of operators are formed, with data streams flowing between them. The code compiles down to an efficient C++ or ML implementation. Our experimentation with WaveScript in VoxNet has shown comparable performance to a hand coded C implementation, and our acoustic localization application has shown a large increase in efficiency and functionality for the end-user.I have contributed various library code to Wavescript, to support our acoustic localization application in VoxNet.
Links: Wavescope wiki