dotnetcpu "Mote": This proof-of-concept sensor network platform is based on the Microsoft SPOT Stamp (dotnetcpu) and the Chipcon CC2420 radio. The dotnetcpu runs the TinyCLR, a small-footprint virtual machine targeted for embedded platforms. One of the great benefits of using the TinyCLR is that Visual Studio can target this virtual machine -- which means programming C# is possible (much easier than C). This platform supports the IEEE 802.15.4 PHY and elements of the MAC. Our C# library implementation includes a simple, UdpClient-like application programming interface, and sample applications for interoperability with IEEE802.15.4 motes that run TinyOS including the Telos and MicaZ. I worked on this project, among other things, during a summer internship at Microsoft Research.

ExScal Mote (XSM): A new sensor network platform for reliably detecting and classifying, and quickly reporting, rare, random, and ephemeral events in a large-scale, long-lived, and retaskable manner. This new mote was designed for the ExScal project which seeks to demonstrate a 10,000 node network capable of discriminating civilians, soldiers and vehicles, spread out over a 10km^2 area, with node lifetimes approaching 1,000 hours of continuous operation on two AA alkaline batteries. This application posed unique functional, usability, scalability, and robustness requirements which could not be met with existing hardware, and therefore motivated the design of a new platform. The detection and classification requirements are met using infrared, magnetic, and acoustic sensors. The infrared and acoustic sensors are designed for low-power continuous operation and include asynchronous processor wakeup circuitry. The usability and scalability requirements are met by minimizing the frequency and cost of human-in-the-loop operations during node deployment, activation, and verification through improvements in the user interface, packaging, and configurability of the platform. Recoverable retasking is addressed by using a grenade timer that periodically forces a system reset. The key contributions of this work are a specific design point and general design methods for building sensor network platforms to detect exceptional events. This project was completed in 2004.

A Line in the Sand: The DARPA-sponsored Line in the Sand ("LITeS") project is a sensor network application designed to provide fine-grained situational awareness in an ad hoc, large, and possibly denied line or area. The network must detect, classify, and track multiple simultaneous targets and their trajectories. This project had many elements - sensing, signal detection, parameter estimation, classification, sensor fusion, tracking, radio communications, time synchronization, routing, and visualization. The sensor nodes combine a processor board, sensor board, a micropower radar, and a power board. These electronic components are housed in a self-righting cylindrical enclosure (versions 1 and 2). The power board was designed for use with the Advantaca micropower radar. The Mica Power Board is a boost regulator that accepts a 3VDC input and provides two output channels that are adjustable from about 3V DC to approximately 12V DC using a pair of onboard potentionmeters. The schematics (OrCAD Capture or PDF), photoplots, BOM, and component datasheets are available online. My contributions to this project included system architecture, signal processing (signal detection, parameter estimation, and elements of classification), electronics hardware design, mechanical hardware design, planning, and project management. This project was completed in 2003.

Cyberstick Pro: The Comdex '97 "Best of Show" awarding-winning CyberStick Pro was the first virtual-reality joystick to use MEMS-based silicon accelerometers (vs. potentiometers) to measure joystick position (in fact, it was the first non-automotive commercial application of the ADXL family (ADXL202) of MEMS accelerometers from Analog Devices). Unlike traditional joysticks, this unit could be held and manuevered in midair. The joystick operated in both analog and digitial modes, and its sensitivity could be adjusted electronically. I was a member of the team that developed this product. This project was completed in 1997.

Wheeled Actively-Articulated Vehicle This six-wheeled mobile robot research platform was built to understand the mobility, coordination, and control issues in actively controlled vehicles. Our testbed was a six-wheeled machine called the Wheeled Actively Articulated Vehicle (WAAV). In contrast with traditional rovers like the NASA/Jet Propulsion Lab Rocker Bogie, the WAAV had twelve actively controlled degrees-of-freedom -- six wheels and six active joints (a pair of roll, pitch, and yaw joints). Our research demonstrated that by actively controlling the normal force between each wheel and the underlying surface, we could significantly improve traction and enhance mobility in unstructured terrain. I worked with S.V. Sreenivasan and Professor Kenneth Waldron on this project. My contributions were design and implementation of several electrical and control subsystems.

Dark Horse: This robot was built for the American Association for Artificial Intelligence/International Joint Conferences on Artificial Intelligence held in Seattle in 1994. The entry won the LEGO Mobile Robot Competition that year after competing against some very strong teams from around the country. The robot had a differential-drive system and a freely rotating caster. It had a horizontally-moving arms that were used to extend its reach to push widgets around. The gear-train is visible in the lower right-hand corner just above the wheel. The robot used shaft encoders for measuring wheel rotations, optical sensors for following lines painted on the contest arena, bump sensors for detecting collisions, motor current sensors for detecting stalls, and a four-way servo-driven rotating turret to angulate a pair of beacons on the course. The robot used a hierarchical behavior-based control program to navigate the course. The robot first attempted a dead-reckoning strategy using only encoders for the control loop, which was fast but not robust, until it detected it was off course (e.g. off the painted path, high motor current draw, bump sensor triggered). Depending on the severity of the problem, the robot would respond differently. A minor error like drifting outside of the lines would trigger only the line-following behavior and the cutover itself appeared transparent. A more severe error like a collision or stall caused the robot to stop moving and begin a recovery procedure during which it would attempt to angulate its position against a pair of beacons on the course and determine its location using an internal course map. Daniel Marcu, Feng Zhao and I designed and built this robot. This project was completed in August, 1994.

Speed Controller and H-bridge Driver: Popular among the members of the Silicon Valley Homebrew Robotics Club, these circuits include a PIC-based speed controller and an H-bridge driver that I designed with Chuck McManis, whose distinguished Silicon Valley career includes being one of the first half-dozen or so people on the Java team at Sun. The circuit drives high-current DC motors (20 to 40 amps, depending on the heat sinks) from servo signals. Chuck maintains a web page that includes a theory of operation, schematics, gerbers, BOMs, and assembly instructions for the speed controller and the H-bridge.

Bamboo iAdvise and Bamboo Broker: This software product consisted of two components - a client piece and a broker piece. The client user would enter personal financial information, goals, preferences, etc. The broker component was essentially an "active" sales force automation tool that would highlight relevant client account activity and enable brokers to make client calls that had a high probability of being converted into revenue generating trades or other sales. I was responsible for developing this product when I worked at Bamboo Systems in Silicon Valley. This project was completed in 1998.