Roadmap for improving code quality

[sibocw]

The NeuroMechFly codebase undergoes constant improvement. This page summarizes the past work and the planned upcoming work:

Version accompanying the NeuroMechFly 2.0 preprint (v0.1.0)

• Implementation of the Gym interface
• Separate repository for the analyses reported in the paper

After NeuroMechFly 2.0 was posted (ordered by priority):

• (Done) Implement major changes to the API that is exposed to the user (v0.2.0)
• (Done) Add tutorials to the documentation website
  • (Done) The code alone from the tutorials are under flygym/mujoco/examples
• (In Progress) Under-the-hood improvement: improve the quality of the code that is not directly exposed to most user so that it is more maintainable. Some of these were actually done before the paper was out.
  • (In Progress) NeuroMechFly (the main simulation class)
    • (Done) Parameters: introduce an additional Parameters class
    • (In Progress) Video rendering: introduce an additional RenderedVideo class to handle most of the visualization logic
    • (Pending) Camera specification
    • (Pending) Refactoring private methods
  • (Done) Vision module: introduce an additional Retina class

In the longer term, we might expand online tutorials to include all analyses conducted for the paper. This means adapting the analysis code to 0.2.x API and moving them from nmf2-paper to flygym/mujoco/examples. We might also work on additional behaviors/analyses, which might lead to version 0.3.x.