Back to projects

Data-Driven Intelligent Guidance Methods

Built neural-network-based fast computation for trajectory and missile reachability, achieving average reachability error below 0.5%.

Neural NetworksParallel Data ProcessingGuidance Intelligence

Context

Conventional reachability computation is costly and often too slow for high-frequency decision loops.

Implementation

  • Established a parallel data generation pipeline for efficient dataset construction.
  • Performed data cleaning and feature transformation for stable model inputs.
  • Trained multiple neural architectures for trajectory and reachability estimation.

Outcome

  • Significantly improved reachability computation speed.
  • Achieved average reachability error below 0.5%, meeting engineering-level accuracy targets.