A survey on evolutionary-aided design in robotics

Abstract

SUMMARYThe evolutionary-aided design process is a method to find solutions to design and optimisation problems. Evolutionary algorithms (EAs) are applied to search for optimal solutions from a solution space that evolves over several generations. EAs have found applications in many areas of robotics. This paper covers the efforts to determine body morphology of robots through evolution and body morphology with the controller of robots or similar creatures through co-evolution. The works are reviewed from the perspective of how different algorithms are applied and includes a brief explanation of how they are implemented.

Co-optimization of Morphology and Behavior of Modular Robots via Hierarchical Deep Reinforcement Learning

“…To avoid sub-optimality and limitations brought by empirical design, there are a growing number of evolutionary and learning-based methods to optimize motion of modular robots [18], [19]. Genetic algorithm (GA) was utilized to search and optimize locomotion gaits of modular robotic systems of the M-Tran, iMobot and UBot robots in various environments [20]- [22].…”

On the Schedule for Morphological Development of Evolved Modular Soft Robots

“…Therefore, the field of Evolutionary Robotics has great motivation to study growth development. Nevertheless, not only is the field focused mostly on evolving the controller without evolving the body [13], but also has development received relatively little attention. Some developmental representations have become popular [14,15], but they have been mostly used for morphogenesis only.…”

Environmental Regulation Using Plasticoding for the Evolution of Robots

“…Within engineering, the research areas related to artificial evolution are that of Evolutionary Computing Smith, 2003, 2015) and Evolutionary Robotics (Nolfi and Floreano, 2000;Doncieux et al, 2015;Nolfi et al, 2016). These fields have addressed the evolution of robot controllers (brains) with considerable success but evolving the morphologies (bodies) has received much less attention (Prabhu et al, 2018). Importantly, the influence of the environment has been even more scarcely investigated.…”

“…First, we allow changes not only to the brain but also to the body. While Evolutionary Robotics systems often focus on the evolution of the brain (Weigmann, 2012;Prabhu et al, 2018), we not only also permit the evolution of the body, but even its development. This is a fundamental refinement, given the known importance the body has for intelligence (Pfeifer and Bongard, 2006).…”