Skip to content

Latest commit

 

History

History
32 lines (22 loc) · 1.48 KB

File metadata and controls

32 lines (22 loc) · 1.48 KB

CE description

Cross-Entropy (CE) model is based on the Reinforcement Learning Cross-Entropy method. This model provides maneuvers table by some inital table through an iterative shifting. The approximate algorithm of such iteration is the following:

  1. generate a sample of random maneuvers tables from the initial table;
  2. evaluate each maneuvers table by the reward at the end of the session;
  3. select some tables with the best reward (using a percentile);
  4. shift the initial table in the direction of the best matrices (an average matrix).

This model has a lot of options (that could be changed during the model training, by the way) and possibilities:

  • to set the first maneuver time - could be "early" or "auto";
  • to set the maneuvers angle relative to the velocity vector - could be "complanar", "collinear" or "auto";
  • to add a reverse maneuver to return to the initial orbit.

CE model has a number of not always pleasant features as well:

  • user-defined maneuvers number;
  • a strong dependence on the initial maneuvers table;
  • a plenty of hyperparameters;
  • a great chance to get into the local optimum.

Also, it should be noticed that CE model could be used for tuning maneuvers table obtained by another models.

Examples: