Challenge

  • Our military client required accurate and timely knowledge about an incoming missile threat and its characteristics during a midcourse discrimination phase;
  • The challenge was to react very accurately in a very short timeline to these incoming threats;
  • There were very many different types of sensors used to extract different features of the threat characteristics producing huge amounts of data which had to be compressed and optimized;
  • Timely and accurate target data had to be relayed to the interceptors with hit-to-kill accuracy in seconds;
  • Rapid and accurate information was especially important during the boost phase of the arsenal threat trajectory;
  • The challenge of midcourse discrimination was to dramatically improve and ensure all sensor data was processed quickly and accurately;
  • Feature extraction and optimization was required to ensure that only true targets were identified to support reliable midcourse intercept of our client's armament.

Our Solution

InferX algorithms improved the speed of processing sensor data by thousands of orders of magnitude. Our solution involved combining two pattern recognition algorithms — one was a Bayesian Network which incorporates complex domain knowledge and can work with incomplete data — the second was a Decision Tree which analyzes data for decision making in a very fast fashion.

How did we do it?

  • We created a hybrid and adaptive "filter" which splits the data between two pattern recognition algorithms;
  • The filter split the data between a Bayesian Network and a Decision Tree algorithms;
  • The Bayesian Net was slower but worked with incomplete data;
  • The Decision Tree was lightning fast but less accurate;
  • Our solution was a trade-off between speed and accuracy to ensure the optimum processing speed.

Results

Our technology improved our client's performance by delivering highly confident results at near-real-time speeds which enabled our client's armament to accurately identify, discriminate, track and destroy an enemy's missile before it reaches its intended target. Our hybrid use of algorithms delivered the right results customized to meet the needs of our military client.