PHYSICS INFORMED NEURAL NETWORK (PINN) FOR TRAJECTORY ESTIMATION OF ARTILLERY SHELLS FROM TARGET LOCATION

Traditional artillery operations rely on pre-computed range tables to determine optimal gun settings for engaging enemy targets. This study presents a novel approach that replaces range tables with inverse optimization of a modified point mass model, leveraging Physics-Informed Neural Networks (PINNs). The proposed neural network estimates aerodynamic coefficients along the trajectory of a 155 mm projectile (modified M107) using only measured trajectory coordinates from a single test fire. This eliminates the need for extensive range tables, resulting in significant cost savings and reduced computational effort. Furthermore, the system predicts gun parameters and projectile trajectory with an error less than or equal to by coupling the modified point mass model with a second PINN trained solely on enemy target location data. This integrated approach enhances targeting precision while offering a cost-effective and efficient alternative to conventional artillery trajectory estimation methods.