ACE: Autonomous Conformity Evaluation of Tensor Data by Means of Novel L1-norm Principal-Component Analysis
Abstract
An autonomous artificially intelligent (AI) machine (or collective of autonomous machines) brings in and runs on a multitude of sensed data. Data quality assurance is a necessary condition to have operational assurance for the autonomous machines. Here, we present a proposal to carry out basic research and develop novel mathematical methods that measure the conformity of each data point with respect to all other collected data in a blind, unsupervised, artificially-intelligent way. The developed mathematical data-conformity evaluation schemes process any given data set represented by a high-dimensional matrix (also known as tensor) and convert each data entry to a continuous zero-to-one "alert conformity value" (zero implying highly conforming data; one implying highly non-conforming data.) Non-conforming sensed data may represent critical, actionable information, e.g. internal system failure, sensor system failure, or external interference and data manipulation. For instance, an unmanned aerial vehicle flight controller gathers information from different sensors (accelerometer, gyro, GPS, barometric pressure, etc.), which are then fused together to make state predictions (e.g. using an enhanced Kalman filter.) Control decisions are then made based on the state predictions. Non-conforming (faulty) sensor data lead to wrong control decisions. By identifying generated non-conforming data, a crash could be potentially prevented. Our basic research plan is to develop new theory and mathematical tools that can highlight non-conforming data values based on recursively refined calculations of L1-norm data subspaces. The calculated (soft zero-to-one) conformity values can then be forwarded to human or machine analysts for appropriate action. The impact of the proposed basic research is broad. Identification of non-conforming data entries will enhance our ability to rapidly identify problems: (i) during testing of new AI autonomous technologies and (ii) during deployment and operation of field-ready AI technologies. The proposed theoretical framework will offer a blind, unsupervised way to identify inappropriate/faulty sensed data independently of the nature of the original data set and for arbitrary data dimensionality (multiway tensors.)
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 09, 2020
- Source ID
- W911NF2010283
Entities
People
- Dimitris A. Pados
Organizations
- Army Contracting Command
- Florida Atlantic University
- Office of the Secretary of Defense