Biomimetic evolutionary analysis: Robotically-simulated vertebrates in a predator-prey ecology

Doorly, Nicole; McArthur, Gianna; Combie, Keon; Engel, V.; Sakhtah, Hassan; Stickles, Elise; Rosenblum, Hannah G.; Gutiérrez, Andrés; Root, Robert G.; Liew, Chun Wai; Long, John H.

doi:10.1109/alife.2009.4937706

Cited by 11 publications

(14 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It has demonstrated an ability to address challenging problems in many task domains including: gaits (Clune et al, 2009), object manipulation (Bongard, 2008), biological study (Crespi et al, 2013;Doorly et al, 2009), and the optimization of morphology (Auerbach and Bongard, 2010;Bongard, 2010;Cheney et al, 2013). Often, these tasks have a single performance objective, or weighted sum, to assess the fitness of each individual.…”

Section: Introductionmentioning

confidence: 99%

Lexicase selection outperforms previous strategies for incremental evolution of virtual creature controllers

Moore¹,

Stanton²

2017

Proceedings of the 14th European Conference on Artificial Life ECAL 2017

View full text Add to dashboard Cite

Evolving robust behaviors for robots has proven to be a challenging problem. Determining how to optimize behavior for a specific instance, while also realizing behaviors that generalize to variations on the problem often requires highly customized algorithms and problem-specific tuning of the evolutionary platform. Algorithms that can realize robust, generalized behavior without this customization are therefore highly desirable. In this paper, we examine the Lexicase selection algorithm as a possible general algorithm for a wall crossing robot task. Previous work has resulted in specialized strategies to evolve robust behaviors for this task. Here, we show that Lexicase selection is not only competitive with these strategies but after parameter tuning, actually exceeds the performance of the specialized algorithms.

show abstract

Section: Introductionmentioning

confidence: 99%

Lexicase selection outperforms previous strategies for incremental evolution of virtual creature controllers

Moore¹,

Stanton²

2017

Proceedings of the 14th European Conference on Artificial Life ECAL 2017

View full text Add to dashboard Cite

show abstract

“…Only the population of Preyros evolved; the single Tadiator was held constant in morphology and coding, providing a consistent force of selection over the generations. Preyro and Tadiator were first introduced in evolutionary experiments by Doorly et al (2009) and their design was elaborated in Long (2012). We briefly summarize and update their design here.…”

Section: Autonomous Robotsmentioning

confidence: 99%

“…But if one seeks to evolve morphology in complex physical circumstances, where no valid physics engine exists, then the new bodies must be fabricated each generation. Thus manufacturing time currently imposes a high labor cost on embodied evolution experiments involving morphology (Long et al, 2006;Doorly et al, 2009;Long, 2012). The experimental testing of physically embodied robots is also time-consuming: robots must be built, maintained, and fixed; a strict protocol must be developed and followed for all procedures; and data from each individual in each experiment must be checked for quality, concatenated with other data, and used appropriately in algorithms for calculating fitnesses and the genotypes of the next generation.…”

Section: Introductionmentioning

confidence: 99%

“…We test this hypothesis using an evolutionary robotic approach: we employ an embodied biorobotic model of an early vertebrate fish species acting as a prey in a predator-prey ecology. The robots, based on the Tadro-class of robots (Long et al, 2006;Doorly et al, 2009;Long et al, 2011a,b;Long, 2012), are behaviorally autonomous, seeking light as a proxy for feeding, while avoiding a pursuit predator.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Testing Biological Hypotheses with Embodied Robots: Adaptations, Accidents, and By-Products in the Evolution of Vertebrates

et al. 2014

Self Cite

View full text Add to dashboard Cite

Evolutionary robotics allows biologists to test hypotheses about extinct animals. In our case, we modeled some of the first vertebrates, jawless fishes, in order to study the evolution of the trait after which vertebrates are named: vertebrae. We tested the hypothesis that vertebrae are an adaptation for enhanced feeding and fleeing performance. We created a population of autonomous embodied robots, Preyro, in which the number of vertebrae, N, were free to evolve. In addition, two other traits, the span of the caudal fin, b, and the predator detection threshold, ζ, a proxy for the lateral line sensory system, were also allowed to evolve. These three traits were chosen because they evolved early in vertebrates , are all potentially important in feeding and fleeing, and vary in form among species. Preyro took on individual identities in a given generation as defined by the population's six diploid genotypes, G i. Each G i was a 3-tuple, with each element an integer specifying N, b, and ζ. The small size of the population allowed for genetic drift to operate in concert with random mutation and mating; the presence of these mechanisms of chance provided an opportunity for N to evolve by accident. The presence of three evolvable traits provided an opportunity for direct selection on b and/or ζ to evolve N as a by-product of trait correlation. In selection trials, different G i embodied in Preyro attempted to feed at a light source and then flee to avoid a predator robot in pursuit.The fitness of each G i was calculated from five different types of performance: speed, acceleration, distance to the light, distance to the predator, and the number of predator escapes initiated. In each generation, we measured the selection differential, the selection gradient, the strength of chance, and the indirect correlation selection gradient. These metrics allowed us to understand the relative contributions of the three mechanisms: direct selection, chance, and indirect selection. Direct selection on N, operating alone, caused the initial increase in N, but was then augmented by chance and indirect selection.Through the course of 11 generations, chance and indirect selection would occasionally supplant direct selection as the primary evolutionary driver. In later generations, direct selection switched sign, stabilizing N at an apparent optimum mean value of 5.7. These results tentatively support the hypothesis that vertebrae evolved as an adaptation for enhanced feeding and fleeing performance in early vertebrates.

show abstract

“…Under selection for enhanced phototaxis, the tail morphology of a population of Tadros evolved (Long et al, 2006). With the addition of a predator, a second eye, and a predator-detection system, constant selection pressure on Tadros for enhanced phototaxis and predator avoidance yielded variable evolutionary patterns, a combination of directional, random, and correlated ("by-product") effects on morphology of the flapping tail and the sensitivity of the predatordetection system (Doorly et al, 2009;Long, 2012;Roberts et al, 2014). In this study, we keep morphology constant and permit the connections of the controller to evolve under selection for enhanced phototaxis.…”

mentioning

confidence: 99%

Modularity and Sparsity: Evolution of Neural Net Controllers in Physically Embodied Robots

et al. 2016

Self Cite

View full text Add to dashboard Cite

While modularity is thought to be central for the evolution of complexity and evolvability, it remains unclear how systems bootstrap themselves into modularity from random or fully integrated starting conditions. Clune et al. (2013) suggested that a positive correlation between sparsity and modularity is the prime cause of this transition. We sought to test the generality of this modularity-sparsity hypothesis by testing it for the first time in physically embodied robots. A population of 10 Tadros -autonomous, surface-swimming robots propelled by a flapping tail -was used. Individuals varied only in the structure of their neural net controller, a 2 × 6 × 2 network with recurrence in the hidden layer. Each of the 60 possible connections was coded in the genome and could achieve one of three states: −1, 0, and 1. Inputs were two light-dependent resistors and outputs were two motor control variables to the flapping tail, one for the frequency of the flapping and the other for the turning offset. Each Tadro was tested separately in a circular tank lit by a single overhead light source. Fitness was the amount of light gathered by a vertically oriented sensor that was disconnected from the controller net. Reproduction was asexual, with the top performer cloned and then all individuals entered into a roulette wheel selection process, with genomes mutated to create the offspring. The starting population of networks was randomly generated. Over 10 generations, the population's mean fitness increased twofold. This evolution occurred in spite of an unintentional integer overflow problem in recurrent nodes in the hidden layer that caused outputs to oscillate. Our investigation of the oscillatory behavior showed that the mutual information of inputs and outputs was sufficient for the reactive behaviors observed. While we had predicted that both modularity and sparsity would follow the same trend as fitness, neither did so. Instead, selection gradients within each generation showed that selection directly targeted sparsity of the connections to the motor outputs. Modularity, while not directly targeted, was correlated with sparsity, and hence was an indirect target of selection, its evolution a "by-product" of its correlation with sparsity.

show abstract

Biomimetic evolutionary analysis: Robotically-simulated vertebrates in a predator-prey ecology

Cited by 11 publications

References 18 publications

Lexicase selection outperforms previous strategies for incremental evolution of virtual creature controllers

Lexicase selection outperforms previous strategies for incremental evolution of virtual creature controllers

Testing Biological Hypotheses with Embodied Robots: Adaptations, Accidents, and By-Products in the Evolution of Vertebrates

Modularity and Sparsity: Evolution of Neural Net Controllers in Physically Embodied Robots

Contact Info

Product

Resources

About