I want to build a scientific aerospace simulator in Nim where the core physics and AI live entirely in Nim, and the system learns the vehicle dynamics instead of hard-coding them. The simulator would model a rocket (later drones), log state and control data, and train a small dynamics model that predicts the next state from the current state and inputs. For accurate, paper-grade 3D projection of trajectories and attitude, I plan to use VTK via Python (PyVista) strictly for visualization, while Nim remains the source of truth for simulation and learning. The trained model would be exported (ONNX or binary) and reloaded into Nim for validation and continued learning. The goal is a clean, research-grade tool, not a game engine. I’m asking whether the Nim community would be interested in helping with performance, math, and ONNX or ML tooling to make this feasible.
what sort of "help" you mean?
1. contribute to your code?
i think not
2. suggest or direct yes, just write your questions. although today AI can answer most of them
I assume you might not have much experience with programming. What is your background, by the way?
I’ve noticed that AI often suggests plans that sound impressive but are unrealistic in practice. It’s usually better to start with something simpler, like a basic flight simulator, and gradually build up.