Auto-Scaling

With the next  MIDACO 6.0 release  an additional multi-objective approach for MIDACO with auto-scaling will be provided. This new approach will be made available additionally to the current user target one and is based on the recently introduced "Utopia-Nadir-Balance" concept described in:

 

Schlueter M., Yam C.H., Watanabe T., Oyama A.

Parallelization Impact on Many-Objective Optimization for Space Trajectory Design  [ PDF ]

International Journal of Machine Learning and Computing (IJMLC), Vol. 6, Issue 1, Pages 9 - 14 (2016)

 

This auto-scaling approach will  fully automatically  create and update weights and utopia / nadir information for a multi-objective optimization problem and strive to find the best equal balanced (or user specified balanced) solution among all objectives. The approach works generally for all pareto front types, in particular convex, concave and separated types. Below is a plot of a concave pareto front with large numerical difference between the two conflicting objectives. The auto-scaling approach delivers fully automatically the best equal balanced solution as final MIDACO solution among the complete pareto front:

( Note the large numerical difference in the X and Y axis values )