Mechanical memory manipulation using Reinforcement Learning

Recent progress has been made in the understanding of memory systems, as they are a promising way to obtain materials with programmable properties. Here we introduce a model framework for dynamical memory manipulation based on a multistable chain composed of coupled bistable spring-mass systems. We show that, using a Reinforcement Learning agent, we can control this highly nonlinear system in force, driving it from any stable or random configuration to any other. We also show that the use of Transfer Learning techniques allows to extend this process to a much larger region of parameter space.