Energid provided methods and software for offloading computationally expensive decisions to remote supervisors. Our focus in this project was on human supervisors, though the techniques apply to remote computers as well. Our approach compensates for shortcomings in fielded hardware by incorporating human intelligence into the tracking process. In tracking systems, many target tracks, both true and false, persist for seconds or even minutes. A remote supervisor can the performance of remote systems by providing intelligent assessment of the quality of these tracks.

The software developed during this project addresses latency and bandwidth limitations. For data transmission, the UDP/IP network protocol is used to transmit information described using the XML framework. Only the components of sensor data relevant to the remote supervisors are transmitted. For example, an entire image will not be transmitted each frame, but rather only the parts containing potential objects of interest. Data is decomposed into segments, with each segment independently compressed. This allows robust supervision in the presence of unreliable communications.

Our techniques enable trackers to perform at the speed of a computer with the intelligence of a human brain. Human input provides levels of performance that will not be possible through on-board computers acting alone for decades. For machine vision, human input provides accuracy and robustness, as humans excel at discriminating targets from background clutter and overcoming anomalies. Human supervisors also support rapid reconfiguration for complex goals. It is easier to convey new, complex information to a human than to rewrite it into machine-vision and control code. In our design, these improvements come with very low risk. When there is no supervision available, the system reverts to traditional behavior.