HARDWARE SYSTEM

 

 

 

TRAJECTORY AND SENSOR DATA FOR EACH NODE

The final stage of this research was to execute a test of the parasitic nodes by releasing them into a real-world situation. More specifically, ten semi-passive, value-added, spherical nodes were given orders and released into an environment populated by human hosts.

The area selected for the test was the third floor of the MIT Media Lab. The floor isusually inhabited by about 40 students/faculty and has light but steady traffic through its pathways. In order to run the test, the floor was first covered with 6 Bluetooth location beacons. With these 6 beacons and the areas where two or three beacons overlap, the floor was divided into a grid of 16 distinct zones.    

          

 

          Distinct zones formed by overlapping beacons

The nodes were then prepared, first by programming the application-specific firmware into each node. This test application firmware uses all three identified states (idle, attached, sensing) and specifically enables and disables peripherals and alters the samplig rates accordingly. This allowed the application to be optimized for power usage.

The nodes were given specific behaviors. Of the ten nodes releasedm, six were told to try and get to specific geographic locations using the distance checking algorithm from the software simulator. The remaining four nodes were given specific conditions to look for. All ten nodes constantly collected sensor data throughout the test, but these four nodes were also programmed with desired sensor conditions (e.g. bright lighting.)

The batteries were completely charged, and the nodes were sealed into their plastic speheres and labeled with instructions. Detachment from the host involved vibrating the pager motor and flashing a red LED in an attempt to be put down.

 

 

 

 

The nodes were then deployed in a high-traffic hallway, and their positions were monitored from any PC on the network. Within 15 minutes, all of the nodes had found their way to a new location. Some were knocked around and rolled, and some were picked up and brought to a new location. The people carrying them mostly obeyed the device when it shook and wanted to be put down, sometimes even tossing it away, startled by the vibration.

The batteries lasted for about 4 hours. The nodes collected data on location, attachment, state changes, sensors, and the time they reached their goals. The battery that was used has a capacity of 145 mAh. The test incorporated many power optimizations but also had many features enabled for logging and observation of the test operation such as the constant flash writes, data dumps over the wireless network, and health monitoring communications with the observer's PC. Further power savings could be achieved by cutting the output power of the radio transceiver from the Bluetooth's specified high power mode to its normal output power mode.

The test generally ran without a hitch, other than the disappearance of one of the ten nodes. This is an expected loss considering the unknowns of the host’s behaviors. The node was found one week later and its sensor data was recovered. Another node got locked inside an office soon after deployment and remained there for the duration of the battery life. One other node was discovered to have manufacturing defests preventing it from recording data to the flash, so it was removed from the test. The remaining eight nodes easily covered the test area. The Figure below shows which node was in which zone at what time.