Log-Based Targeting

A family of targeting strategies that store past data about enemy movement in an unprocessed or preprocessed form. These methods often perform extensive calculations on the entire data set in order to deduce a firing angle at fire time; this is in contrast to statistical targeting methods, which generally process input on the fly and simply query an internal data structure to generate a firing angle.

Forms of log-based targeting

• Pattern Matching - The most common form of log-based targeting, this method takes the most recent series of enemy movements and tries to find a previous series where the enemy moved in that pattern.
• Dynamic Clustering - Pioneered by ABC in TronsGun, this system looks through previous scans of the enemy to find some n most similar scans and extrapolates a firing angle from that cluster of scans. It is similar to a k-nearest neighbor algorithm.

See also

• State Matching