SALSA

SALSA

Overview of SALSA (System using Artificial Life to Study Adaptation)

SALSA is an artificial life program. It simulates a two-dimensional world, creatures that move around in that world ("critters"), and the food and rocks that they encounter. Critters have the potential to learn how to negotiate their world since they receive rewards and punishments for their actions. Their "brains" use reinforcement learning to allow them to behave more appropriately to whatever environment they have been placed in.

Each critter has an array of sensors (for gathering input from the world) and effectors (for producing actions in the world). Sensors give the critters vision, sonar and touch. Effectors give critters the ability to move, eat, and attack. Some sensors and effectors are more appropriate for some tasks than others. For instance, in a world where plants are abundant, critters may not need to move very fast; they can reach food with small steps. In a world where all of the other objects are the same color, color vision may not be very useful. The proper combination of sensors and effectors can be found only through an analysis of the environment and actual experimentation with the world, though some tasks may be solved by multiple combinations of sensors and effectors.

In addition to learning about reinforcement learning, this program demonstrates the importance of the physical body (embodiment) and being immersed in a physical environment (situatedness). Although it might seem that a big, adaptible brain is all an organism needs to solve any problems it encounters, the type and quality of sensors it has affects what it can know about its world. The same is true for the organism's effectors (things that allow the organism to act in the world). In addition, there is no perfect sensor or effector. In some environments, the "best" eye can be useless (e.g., under the earth or deep in the ocean, light is virtually nonexistent), and the best ear might be a distraction. Each organ is better or worse depending on how it fits with its environment. This is even true of brains. A big brain can be useless to an animal that doesn't need complex behavior, and it might even be detrimental; all brains require large amounts of space and energy.

That said, sometimes a big brain, good sensors and effectors, and a powerful learning mechanism are needed in order to do well in some environment. By experimenting with different environments and critters in SALSA, you will see that sometimes the simplest creature is the best, and sometimes a more complex creature is the only one capable of surviving in its world. You will even see that some critters do well in many worlds, while others are limited to just a few specific ones.

Finally, SALSA is a simulation of a large number of interacting individuals, or agents. Although individual actions are important, population-level phenomena can also be studied with this simulation. Critters in Salsa will demonstrate mass extinction (sadly, a common occurrence), competition, and cooperation. Often, it is the properties not only of the world but of the group that will affect the success of a particular critter or group of critters.

The SALSA Simulation