morphological_novelty_metric

Calculate Morphological Novelty across a Population.

Using the MorphologicalNoveltyMetric for evaluating a population of robots can be done as shown in the code below.

>>> robots: list[revolve2.modular_robot.ModularRobot]
>>> novelties = get_novelty_from_population(robots)

Submodules

Package Contents

Functions

get_novelty_from_population(...)

Get the morphological novelty score for individuals in a population.
get_novelty_from_population(population: list[revolve2.modular_robot.ModularRobot], cob_heuristic: bool = False, num_bins: int = 20) → numpy.typing.NDArray[numpy.float64]

Get the morphological novelty score for individuals in a population.

This metric for Morphological Novelty considers robots a distribution in space, which can be reshaped into any other distribution. The work that has to be done to reshape distribution 1 to distribution 2 is used for the final novelty calculation. A detailed description of the Algorithm can be found in: Oliver Weissl, and A.E. Eiben. “Morphological-Novelty in Modular Robot Evolution”. 2023 IEEE Symposium Series on Computational Intelligence (SSCI)(pp. 1066-1071). IEEE, 2023.

Parameters:

  • population – The population of robots.

  • cob_heuristic – Whether the heuristic approximation for change of basis is used.

  • num_bins – The amount of bins in the histogram. Increasing this allows for more detail, but risks sparseness, while lower values generalize more.

Returns:

The novelty scores.