Modular neuroevolution for multilegged locomotion

Valsalam, Vinod K.; Miikkulainen, Risto

doi:10.1145/1389095.1389136

Cited by 43 publications

(38 citation statements)

References 25 publications

(32 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Evolutionary computation, frequently involving the evolution of neural network controllers, has been successfully used to this end [3]- [10]. Evolved gaits are often better than those produced by human designers; one was even included on the commercial release of Sony's AIBO robotic dog [3], [7]. However, many researchers have found that they cannot hand the entire problem over to evolutionary algorithms, because of the large number of parameters that need to be simultaneously tuned to achieve success [3], [7], [9]- [15].…”

Section: Introductionmentioning

confidence: 99%

“…Evolved gaits are often better than those produced by human designers; one was even included on the commercial release of Sony's AIBO robotic dog [3], [7]. However, many researchers have found that they cannot hand the entire problem over to evolutionary algorithms, because of the large number of parameters that need to be simultaneously tuned to achieve success [3], [7], [9]- [15]. Many of these scientists report that, while it is possible to evolve a controller to manage the inputs and outputs for a single leg, once evolution is challenged with the inputs and outputs of many legs, it fails to make progress.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Evolving coordinated quadruped gaits with the HyperNEAT generative encoding

Clune

Beckmann

Ofria

et al. 2009

2009 IEEE Congress on Evolutionary Computation

146

187

View full text Add to dashboard Cite

Abstract-Legged robots show promise for complex mobility tasks, such as navigating rough terrain, but the design of their control software is both challenging and laborious. Traditional evolutionary algorithms can produce these controllers, but require manual decomposition or other problem simplification because conventionally-used direct encodings have trouble taking advantage of a problem's regularities and symmetries. Such active intervention is time consuming, limits the range of potential solutions, and requires the user to possess a deep understanding of the problem's structure. This paper demonstrates that HyperNEAT, a new and promising generative encoding for evolving neural networks, can evolve quadruped gaits without an engineer manually decomposing the problem. Analyses suggest that HyperNEAT is successful because it employs a generative encoding that can more easily reuse phenotypic modules. It is also one of the first neuroevolutionary algorithms that exploits a problem's geometric symmetries, which may aid its performance. We compare HyperNEAT to FT-NEAT, a direct encoding control, and find that HyperNEAT is able to evolve impressive quadruped gaits and vastly outperforms FT-NEAT. Comparative analyses reveal that HyperNEAT individuals are more holistically affected by genetic operators, resulting in better leg coordination. Overall, the results suggest that HyperNEAT is a powerful algorithm for evolving control systems for complex, yet regular, devices, such as robots.

show abstract

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Evolving coordinated quadruped gaits with the HyperNEAT generative encoding

Clune

Beckmann

Ofria

et al. 2009

2009 IEEE Congress on Evolutionary Computation

146

187

View full text Add to dashboard Cite

show abstract

“…The evolution of locomotion gaits for multilegged robots is a classical problem in EC, addressed in many prior studies utilizing both direct and generative encodings (e.g., Liu and Iba (2004); Clune et al (2011);Valsalam and Miikkulainen (2008)). In the large majority of these studies, the performance of evolved individuals is analyzed solely by the walking speed of the robot and the required number of generations of evolution.…”

Section: Hexapod Locomotion Problemmentioning

confidence: 99%

Comparing the Evolvability of Generative Encoding Schemes

Tarapore

Mouret

2014

Artificial Life 14: Proceedings of the Fourteenth International Conference on the Synthesis and Simulation of Living Systems

View full text Add to dashboard Cite

The evolvability of a system is the ability to generate heritable, novel and non-lethal phenotypes, from random genetic mutations. However, most evolutionary computation studies estimate evolvability either as, (i) the proportion of mutations beneficial to an individual's performance, irrespective of the phenotypic diversity of the mutants, or (ii) the range and diverseness of mutated phenotypes, without taking into account the viability of the genetic change. This paper reports a novel approach to measure the evolvability provided by an encoding, by characterizing both the quality of the mutations and the quantity of phenotypic variation. We evolved controllers for hexapod robot locomotion using a parameterized direct encoding, and the generative encoding of artificial neural networks (similar to HyperNEAT) and single-unit pattern generators (SUPGs). Our results reveal that the performance of an encoding is not always a good assessment of evolvability. Although both the generative encodings evaluated had individuals with high performance gaits, there were apparent differences in their measured evolvability. A direct and predictive relationship is indicated between our measure of evolvability, and the number of generations required by amputated individuals to recover an effective gait.

show abstract

“…The controller for a multilegged robot can be implemented as a system of interconnected neural network modules, each controlling a different leg [1,34]. Some of these modules and interconnections may be identical, resulting in symmetries, i.e.…”

Section: The Enso Approachmentioning

confidence: 99%

Constructing controllers for physical multilegged robots using the ENSO neuroevolution approach

et al. 2012

Self Cite

View full text Add to dashboard Cite

Evolving controllers for multilegged robots in simulation is convenient and flexible, making it possible to prototype ideas rapidly. However, transferring the resulting controllers to physical robots is challenging because it is difficult to simulate realworld complexities with sufficient accuracy. This paper bridges this gap by utilizing the Evolution of Network Symmetry and mOdularity (ENSO) approach to evolve modular neural network controllers that are robust to discrepancies between simulation and reality. This approach was evaluated by building a physical quadruped robot and by evolving controllers for it in simulation. An approximate model of the robot and its environment was built in a physical simulation and uncertainties in the real world were modeled as noise. The resulting controllers produced well-synchronized trot gaits when they were transferred to the physical robot, even on different walking surfaces. In contrast to a hand-designed PID controller, the evolved controllers also generalized well to changes in experimental conditions such as loss of voltage and were more robust against faults such as loss of a leg, making them strong candidates for real-world applications.

show abstract

Modular neuroevolution for multilegged locomotion

Cited by 43 publications

References 25 publications

Evolving coordinated quadruped gaits with the HyperNEAT generative encoding

Evolving coordinated quadruped gaits with the HyperNEAT generative encoding

Comparing the Evolvability of Generative Encoding Schemes

Constructing controllers for physical multilegged robots using the ENSO neuroevolution approach

Contact Info

Product

Resources

About