Evolving legged robots using biologically inspired optimization strategies

Abstract

Abstract-When designing a legged robot a small change in one variable can have a significant effect on a number of the robot's characteristics, meaning that making tradeoffs can be difficult. The algorithm presented in this paper uses biologically inspired optimization techniques to identify the effects of changing various robot design variables and determine if there are any general rules which can be applied to the design of a legged robot. Designs produced by this simulation are also compared to existing robot designs and biological systems, showing that the algorithm produces results which require less power than other robots of a similar mass, and which share a number of characteristics with biological systems.

A survey on evolutionary-aided design in robotics

“…While most populations were initiated with random values, Samuelsen et al [34] chose 200 identical individuals and the GOLEM project [73] had 200 to 1000 null individuals to start the evolutionary process. Single population evolved solutions except in [74,75]. In [74], for faster convergence, when the average fitness of the first population reached a fixed value (or after 50 generations), the best individual was transferred to a second population.…”

“…Single population evolved solutions except in [74,75]. In [74], for faster convergence, when the average fitness of the first population reached a fixed value (or after 50 generations), the best individual was transferred to a second population. Once twenty such transfers were completed, the solutions were further optimised with the second population.…”

“…Truncation strategy appeared in [34], and elitism with tournament selection was applied by Clark et al [68] and Moore et al [36]. Researchers [50] replaced the loser of the population with the child and in [74], individuals with costs higher than average of the population were replaced. Further, [44] retained random non-dominated solutions by replacing dominated solutions and parents were always replaced in [37].…”

“…SGAs possess flaws, to achieve faster convergence, they have been modified to include elitism and tournament selection [33]. SGAs were used for morphology development in [50,56,63,66,74,83,84] and they were amended with elitism, or other selection methods [35,38,43,45,46,48,51,68,87,88]. A GA with novelty search and an objective based GA found application in [70] where individuals that performed differently were rewarded if they possessed novelty.…”

Novelty-Based Evolutionary Design of Morphing Underwater Robots

“…switching from running to swimming to overcome an obstacle). Also, with respect to previous works following similar approaches [25] (focusing on the evolution of quasi-static gaits for terrestrial robots), the setting presented in this paper differs for several reasons. First, robots will evolve in water environment, in which, with respect to the terrestrial one, the importance of morphology is further stressed.…”