A Hormone-Driven Epigenetic Mechanism for Adaptation in Autonomous Robots

Abstract

Different epigenetic mechanisms provide biological organisms with the ability to adjust their physiology and/or morphology and adapt to a wide range of challenges posed by their environments. In particular, one type of epigenetic process, in which hormone concentrations are linked to the regulation of hormone receptors, has been shown to have implications for behavioral development. In this paper, taking inspiration from these biological processes, we investigate whether an epigenetic model based on the concept of hormonal regulation of receptors can provide a similarly robust and general adaptive mechanism for autonomous robots. We have implemented our model using a Koala robot, and tested it in a series of experiments in six different environments with varying challenges to negotiate. Our results, including the emergence of varied behaviors that permit the robot to exploit its current environment, demonstrate the potential of our epigenetic model as a general mechanism for adaptation in autonomous robots.

Autonomous Reinforcement Learning of Multiple Interrelated Tasks

“…I N recent years, the development of autonomous agents has gained increasing interest in the fields of artificial intelligence, robotics, and machine learning. Although autonomy and versatility are pursued via different kinds of approaches such as information theory [1], [2], evolutionary computation [3], deep learning [4], or epigenetic models [5], the field of Corresponding author: V.G. Santucci, vieri.santucci@istc.cnr.it developmental robotics [6], and in particular the research…”

A Robot Model of Stress-Induced Compulsive Behavior

“…First, stress hormone is released from an internal source in proportion to the level of the three error-signaling hormones. This is similar to the stress mechanism in [11]. Second, stress hormone is released due to an external source when the robot is physically confined.…”

“…Corresponding to each physiological variable is a simulated hormone that signals the deficit (or error) of that variable, as in [11]. The hormone is released at a rate proportional to the current deficit (the difference between the current value of the variable and the robot's target for that variable) and decays exponentially at a rate proportional to the current level of the hormone.…”

Embodied Robot Models for Interdisciplinary Emotion Research

“…Models of neuromodulation in robots span different aspects, such as increasing the plasticity and flexibility of the robot controller [28], [90], improving its evolvability [33], [78], improving its adaptation and survival by affecting its action selection [38], [56], producing different emotionrelated behaviors (different functionalities) from the same underlying controller [4], [22] or affecting its emotional and cognitive development [11], [65], [66]. Let us have a look at some of the models developed in my group that implement emotions as "emergent functionality" [91], relating them to the consequences and guidelines of the neuroscience model proposed by Fellous.…”