Robotics Technology Center

Development of Adaptive Behavior in Robots

There is a new experiment done by some researchers. They developed small, mobile robots that are controlled by onboard neural networks capable of learning.

Multiple experiments are done and they found out that robots evolved complex behavioral patters such as homing, co-adaptation of brains and bodies, sophisticated predator versus prey strategies, collision-free movement, cooperation, and even altruism. This experiment was done by the means of Darwinian selection.

Example of an evolved “creature” created by autonomous design and fabrication process. Image: Hod Lipson

First step of the experiment was to carried out real robots allowing the researchers to show a video about the behaviors of the evolved robot.

Second, the neural architecture of the robot is a simple architecture with no ontogenetic development, no synaptic plasticity and no detailed modelling of ion channels and spike transmission.

Third, the genomes were directly mapped into the neural network for example, time- dependent dynamics, ontogenetic plasticity, or gene-to-gene interaction.

According to the researchers, by narrowing their analysis to these studies, they are able to emphasize the strength of the process of Darwinian selection in comparable simple systems exposed to distinct environmental conditions.

This experimental evolution with robots proves the power of evolution by recombination, alteration and natural selection.

Click here to read more about Darwinian Selection in Robots, Natural Adaptive Behavior in Robots