We're excited to share the deep dive into our overhauled modeling and simulation engine at Collimator. Our engine brings together the best of Python, JAX, Drake, and cloud technology to optimize engineering workflows, particularly in aerospace and robotics.
In the article, we cover:
- The rationale behind our choices
- How our engine leverages JAX for performance and automatic differentiation
- The integration of machine learning with traditional physics-based modeling
- Advanced optimization techniques
- Real-world applications and examples
At Collimator.ai we built a differentiable modeling and simulation platform for dynamical systems on JAX (think Simulink but in the cloud and with Python+JAX instead of Matlab) in order to enable more efficient optimization, autotuning, MPC, neural network control, etc. Here is a simple example of PID tuning with JAX and autodiff: https://py.collimator.ai/examples/pid_tuning/ (feedback is welcome!)
In the article, we cover: - The rationale behind our choices - How our engine leverages JAX for performance and automatic differentiation - The integration of machine learning with traditional physics-based modeling - Advanced optimization techniques - Real-world applications and examples
We'd love to hear your thoughts and feedback!