Environmental Adaptation of Robot Morphology and Control Through Real-World Evolution

Abstract

Robots operating in the real world will experience a range of different environments and tasks. It is essential for the robot to have the ability to adapt to its surroundings to work efficiently in changing conditions. Evolutionary robotics aims to solve this by optimizing both the control and body (morphology) of a robot, allowing adaptation to internal, as well as external factors. Most work in this field has been done in physics simulators, which are relatively simple and not able to replicate the richness of interactions found in the real world. Solutions that rely on the complex interplay between control, body, and environment are therefore rarely found. In this paper, we rely solely on real-world evaluations and apply evolutionary search to yield combinations of morphology and control for our mechanically self-reconfiguring quadruped robot. We evolve solutions on two distinct physical surfaces and analyze the results in terms of both control and morphology. We then transition to two previously unseen surfaces to demonstrate the generality of our method. We find that the evolutionary search finds high-performing and diverse morphology-controller configurations by adapting both control and body to the different properties of the physical environments. We additionally find that morphology and control vary with statistical significance between the environments. Moreover, we observe that our method allows for morphology and control parameters to transfer to previously-unseen terrains, demonstrating the generality of our approach.

Morpho Evolution With Learning Using a Controller Archive as an Inheritance Mechanism

“…For example, this excludes work in which morphological change is restricted to repositioning sensors [11] or altering the length, weight and size of joints, e.g. [12], [13], [14], [15]. We first discuss methods that create controllers that are directly correlated to a specific type of body-plan, followed by morphology-independent methods, i.e.…”

Real-world embodied AI through a morphologically adaptive quadruped robot

“…The robot is powered by a three cell 5Ah LiPo battery supplying the robot with an unregulated voltage of 12.6V when fully charged. It has previously been used in laboratory settings, e.g., [9,52].…”

“…1a. The quadruped robot, DyRET (the Dynamic Robot for Embodied Testing), provides a powerful proof of concept harnessing a variable morphology to adapt to realistic real-world conditions in outdoor settings [8,9]. Morphological adaptation is provided through variable-length legs, whereby the length of both femur and tibia can be adjusted on the fly to enable different walking behaviours whilst also tilting the central body (Fig.…”

“…Adaptation can be realized purely through software, primarily adapting gait patterns and foot-tip arcs. Techniques that allow locomotion on challenging terrain include evolutionary approaches [9,10,11], reinforcement learning [12,13], and Bayesian optimization [14,15], as well as perception-less [16] and hybrid approaches [17,18,19]. Online adaptation to terrains of different compliance under aggressive maneuvers and external disturbances has been studied [20], as well as walking posture adaptation for navigation in confined spaces [21].…”

EMERGE Modular Robot: A Tool for Fast Deployment of Evolved Robots

“…To the best of our knowledge, the full cycle evolution without human intervention, i.e. an evolutionary process in hardware that is capable of configuring, testing, and reconfiguring robots to start over again, has only been implemented for morphological variations in the length of the limbs of a robot ( Nygaard et al, 2020b ). Until now, it has not been achieved for other 3D morphologies that can more drastically reconfigure the arrangement of the robot parts, that is, that have more topological variability.…”