Robotic evolution affected by both nature and nurture
A robot‘s experience, as well as its environment and its ‘genome’, has been shown to affect its evolution, making the field of evolutionary robotics far more complex and interesting than previously thought.
A team of New York-based researchers studied the evolution of embodied robots over ten generations, as robots completed tasks to test their fitness, and found that developmental factors affected how the population evolved.
This is the first time researchers have taken epigenetic factors into account in embodied robot evolution.
Evolutionary robotics is a developing field which applies the principles of natural selection to allow autonomous robots to develop new traits without them having been explicitly programmed. Earlier this year, a six-legged robot developed a gait never observed before in nature after scientists ran an evolutionary-like algorithm to optimise its speed.
Embodied robots undergo a parallel to Darwinian evolution when an artificial ‘gene pool’ is created, with genomes which determine how a robot operates. Normally, the robots are made to perform tasks and their fitness is ranked according to how well they complete them. By combining their genomes, the fittest robots ‘reproduce’.
However, biological evolution is not as simple as Darwin envisaged. It has been established that our genomes are altered by our environment and experiences, not just by random mutations. For instance, a person who frequently engages in addictive behaviours such as smoking or gambling risks passing these addictive epigenetic traits down to their descendants. This interplay between evolution and development is referred to as ‘evo-devo’.
This study, published in Frontiers in Robotics and AI, suggests that epigenetics is significant not just in biology, but also in robotics.
“An explicit evo-devo approach has proven invaluable in the evolution of artificial neural networks,” say Jake Brewer and Aaron Hill, of Vassar College in New York, USA. “Development serves as a new type of evolutionary driver – alongside the genetic factors of mutation, recombination and selection.”
“We note that what is missing from evolutionary robotics is not development per se but rather physically embodied development. We take the first simple steps toward combining the two by examining the interactions of epigenetic and genetic factors in the evolution of physically embodied and simulated robots.”
The robotic ‘genome’ was a binary code which determined how their hardware should be wired. Populations of robots completed tasks to determine their fitness; light gathering and obstacle avoidance. The robots were then randomly mated by using an algorithm to combine genomes, and the next generation of robots were rewired according to their new genomes. There were two groups of robots; a group for which epigenetic factors were accounted for, and a control group. This allowed the researchers to explore a robotic analogue for epigenetics for the first time.
Over ten generations, the researchers observed the evolution of the robots, and found that the group of robots for which epigenetic factors were accounted for underwent significantly different evolution to the control group.
“To our knowledge, our work represents the first physically embodied epigenetic factors to be used in the evolution of physically embodied robots,” Brawar and Hill said.