ALPACA, Adaptive Learned Priority Anytime Coding Algorithm
Department of Defense
Agency Tracking Number:
Solicitation Topic Code:
Small Business Information
Charles River Analytics Inc.
625 Mount Auburn Street, Cambridge, MA, 02138
Socially and Economically Disadvantaged:
AbstractFuture aircraft carriers will rely increasingly on automation as manning levels continue to decline, and opportunities are identified for “smart” automation. One particular area with great potential is in the health monitoring and diagnostics of aircraft launch and recovery equipment, through reachback to on-shore services. A key technical hurdle in realizing this capability is achieving adequate data compression to accommodate low- and variable-bandwidth channels: existing compression algorithms simply do not meet the challenge because they are either lossless, and therefore not adaptive to low bandwidth, or are based on perceptual models, and not necessarily relevant to diagnostic tasks. However, by “learning” the relevance of on-board data to the on-shore diagnostic task, we can prioritize the most relevant information to be transmitted quickly, with more complete data to follow. We also develop anytime codes to address adapting to bandwidth. We propose ALPACA, Adaptive Learned Priority Anytime Coding Algorithm, which combines a distortion function learned from the on-board data to prioritize information in the encoding, with anytime coding, a novel technique for sending a message that can be reliably decoded if truncated. ALPACA is optimized for transmitting in a low- and variable-bandwidth setting, enabling rapid decision-making for remote diagnostics of equipment.
* information listed above is at the time of submission.