nmboyd / g-fold-python

Rocket Landing "Guidance for Fuel Optimal Diverts" in Python

G-FOLD Python

Guidance for Fuel Optimal Landing Diverts (GFOLD)

This code reimplements GFOLD algorithm in Python with use of the fantastic cvxpy utility. The algorithm was defined by a number of papers, but chiefly this paper by Ackimese, Carson, and Blackmore at JPL.

What you can do with GFOLD-Python

• Calculate fuel optimal spacecraft landing trajectories
• Generate embeddable C code for real-time trajectory calculation (~0.3 second calculation time 1x 2.4GHz)
• Autonomous landing of Kerbal Space Program rockets

What you can't do with GFOLD-Python

• Attitude control
• Robust control
• Control of any kind (this is a guidance algorithm!)

How to use it

• If you wish to do a static calculation (not generating C code)

  1. Define your vehicle and environment in GFOLD_Static_Parms
  • All notation follows conventions laid out in the original paper
  2. Comment / Uncomment the constraints you wish to have in GFOLD_Static
  3. Run python GFOLD_Static.py (requires scipy)
  4. View the "evil" plots (this name is just a joke btw)

• If you wish to do C code generation

  1. Set test = 0 at the top of GFOLD_Generate.py
  2. Run python GFOLD_Generate.py (requires cvxpy_codegen)
  3. Fix some of the known-bugs cvxpy_codegen creates - See issues page of the repo - Attempt to compile, and solve each error as they come
  4. Compile the generated C code
  5. (Optional:) Install the compiled CPython code into your Python distribution with setup.py if you wish to use the compiled code from Python - Be aware that the Python2.7 Windows Compiler provided by Microsoft will not work because it has a pathetically tiny stack heap size. Recommend using MinGW on Windows!

Documentation

• Since this is a pre-alpha research project, the main documentation is found in #code comments, and in the content of the paper itself.

Requirements

• Python 2.7 (I'm sorry about still using python 2, Mr. Guido, but cvxpy_codegen is the constraint here...)
• scipy (for static solutions)
• cvxpy (for static solutions)
• cvxpy_codegen (for code generation)

License

GPLv3, copyleft license.

Chose this license because I spent way too many late nights and heartbeats working on this - and want to see what people do with it and have changes propagated forward!