Analysis & Assessment
RAM Laboratories is providing analysis and assessment technologies and services that support tracking, trend identification, pattern recognition, and anomaly detection within large volumes of data. Our solutions consider disparate data sources to include historical databases, near-real-time C4I data and real-time sensor data. Our approaches include support for the following:
Trend Analysis and Prediction
Multi-Level Data Fusion
- Air Force Research Laboratory
- Edwards Air Force Base / Air Force Flight Test Center
- Missile Defense Agency
- US Army Space and Missile Defense Command
- US Army Research Development and Engineering Command
Trend Analysis and Prediction: Our scientists and engineers employ a variety of techniques to assess and predict trends in historical and/or real-time data. Our approaches include the use of Modeling and Simulation, Kalman Filtering, and the use of soft-computing approaches (neural networks, genetic algorithms, fuzzy sets) to intelligently assess trends and patterns in data.
Kalman Filters: Kalman Filters are employed to provide both prediction and estimation capabilities based on sensor data sources. Our Kalman Filter implementations go beyond standard approaches and incorporate neural networks and genetic algorithms to address the tracking of highly dynamic systems. Similar solutions have been used by RAM Laboratories to address the tracking of multiple targets that exhibit highly non-linear motion.
Neural Networks: Neural Networks are used for data classification and pattern detection over a wide variety of data sets. Our scientists implement neural networks to address the detection or characterization of data residing in large data volumes while working to reduce training and maintenance times. Neural networks have been used by RAM Laboratories to detect malicious intruders in computer networks, detect anomalies in satellite telemetry data, and to augment traditional Kalman Filter-based tracking approaches.
Genetic Algorithms: Planning activities must consider available resources, approaches, decision points, and objectives. Identifying optimal plans requires the use of techniques to align resources and apply certain actions to a problem in the proper sequence. Our engineers are applying genetic algorithms to address multi-objective optimization in support of planning processes with regard to C2 and adversary analysis.
Bayesian Estimation: Our engineers and scientists use Bayesian estimation to assess the probabilities of certain events occurring RAM Laboratories’ projects have used Bayesian estimation for model selection and multi-modeling, and cognitive modeling when developing rule-based systems to reduce the cognitive workload of system users.
Cognitive Modeling: RAM Laboratories implements cognitive models to provide intelligent computing approaches to analysis and assessment activities, while reducing cognitive workload of system operators. RAM Laboratories' approaches, implemented as Bayesian networks, rule bases, fuzzy sets, neural networks and genetic algorithms, serve to guide Commanders and operators on sensing trends in data.
Multi-Level Data Fusion: Our engineers and scientists employ multi-level techniques that fuse data from a variety of sources. Our techniques implement Level 0 detection algorithms, Level 1 tracking algorithms, Level 2 correlation approaches, and Level 3 characterization and response to a variety of events.