n-body playground

Requires Python 3.5+. License: MIT

Components

• Initial body config generator (from oscullating elements or state vectors);
• Main simulation;
• Analysis toolkit.

Simulation engines and methods

• dummy: for testing only

  • dummy_method

• naive: straightforward Python implementation without massive Numpy/Scipy usage

  • semi_explicit_euler
  • explicit_rk2
  • explicit_rk4
  • ralston
  • bogacki_shampine (TODO)
  • dormand_prince (TODO)
  • explicit_rk16 (TODO)
  • crank_nicolson (TODO)
  • adams_bashforth_2
  • adams_bashforth_5
  • velocity_verlet
  • ruth3
  • ruth4

• scipy: uses scipy.integrate.solve_ivp
• numpy (TODO): massive Numpy/Scipy usage
• cython (TODO): cythonized integration loop
• numba (TODO)
• cpp (TODO): main loop in C++

Usage

Simulation parameters are set in a single YAML config file. See sample_configs/ for examples and description (TODO). Simulation name is the stem (filename without extension) of the config file, it is used in the intermediate file name generation.

To run simulation:

1. Create YAML config (e.g. simulation.yaml)
2. Run prepare_config.py, which will generate simulation.global and simulation.bodies.
3. Run nbody.py (using simulation name as a parameter, here simulation), which will output to a directory simulation/.
4. Do some analysis using plot.py (using simulation name as a parameter).

Issues and roadmap

• Gather accuracy info: build energy error graphs for sample_configs/oscullating.yaml for different dt and methods.
• Gather performance info: create a table with per-iteration speeds for each method.
• Add plotting of effective oscullating elements for each body
• Upgrade position plot to use oscullating element "smoothing" to produce nice plots when T is large but output_point_num is low.
• More methods for naive engine.
• Create a list of useful methods for other engines.
• How numpy can be implemented?
• Implement cython engine.
• Better documentation
• Tests