A screenshot of two Chromaria worlds.

Chromaria. Each creature is born at the center of the world (left) and then must find an appropriate place to plant. The color-rich borders initially provide the only viable options, but more emerge as Chromarians continue to plant in the environment (right).” Figure and caption from .

Details of System

From the first paper on Chromaria (citations removed for clarity):

The world introduced in this paper, called Chromaria, is visually two-dimensional and composed of discrete RGB pixels [figure above]. The colorful creatures (called Chromarians) evolved in this world actively explore it to search for a place to plant. Each Chromarian is allowed one planting attempt. If the Chromarian’s RGB sensor field (which can sense prior successful planters and the background) satisfies a specific planting function involving matching its color (detailed later), then the planting attempt succeeds and the successful creature is eventually allowed to reproduce. Thus the MC in Chromaria, unlike Earth’s MC, is to navigate to a position in the world with colors matching the Chromarian’s own coloring. If this MC is not met, then the Chromarian is removed without planting and does not reproduce.

Each Chromarian’s morphology consists of a twodimensional image composed of RGB pixels. The genetic encoding of this morphology is a compositional pattern producing network (CPPN), a neural-networklike representation that generates patterns with regularities such as symmetry, repetition, and repetition with variation. The CPPN used to encode Chromarian morphologies, which is similar to the encoding in Risi et al. (2012), takes polar coordinates r and θ as input. Such polar coordinates define an unambiguous solid border for the body, which would be harder to determine if the inputs were Cartesian. Upon activation, the CPPN returns an rmax for each value of θ, which determines the perimeter of the Chromarian’s body at that angle. Then every pixel on the interior of this border is queried by the CPPN for the corresponding RGB values at the queried (r, θ), where r is scaled from [0,49] to [0,1], and θ from [−Π,Π] to [-1,1]. In this waythe CPPN determines both the shape (via the rmax output) and internal color (via the RGB outputs) of the Chromarian. These characteristics ultimately determine where the Chromarian can successfully plant. By evolving new colors, Chromarians in effect create novel opportunities for new kinds of planters, thereby satisfying Condition 2.

Each Chromarian is equipped with a 10 × 10 rectangular sensor field that perceives the RGB values (each scaled from the range [0, 255] to [−1, 1]) of the underlying pixels. This field is centered at the forefront of the Chromarian’s body, with half of the pixels falling underneath the body and the rest extending in front of the creature. The exact resolution of the field depends on the creature’s morphology; as its length and width increase, the distance between neighboring sensors grows. Note that Chromarians can overlap if they have planted in the same location, in which case the pixels of the most recent planter are sensed. Additionally, each Chromarian is equipped with a heading-sensitive compass consisting of 8 pie slice sensors. All sensors are input to a multimodal neural controller, whose weights are encoded using a second CPPN following the HyperNEAT approach to encoding large-scale ANNs with CPPNs. The output layer, which receives connections from the hidden layers, has four effector nodes corresponding to the Chromarian’s requested rotation (L and R), speed (S), and desire to plant itself (P). If the planting node exceeds a threshold, then the Chromarian is immobilized and it never moves again. Otherwise, the rotation and speed nodes determine the Chromarian’s next movement.


Soros, L., & Stanley, K. (2014). Identifying necessary conditions for open-ended evolution through the artificial life world of Chromaria. In H. Sayama, J. Rieffel, S. Risi, R. Doursat, & H. Lipson (Eds.), ALIFE 14: Proceedings of the Fourteenth International Conference on the Synthesis and Simulation of Living Systems (pp. 793–800). MIT Press. https://doi.org/10.7551/978-0-262-32621-6-ch128