Maze exploration behaviors using an integrated evolutionary robotics environment

Nelson, Andrew; Grant, Edward; Galeotti, John; Rhody, Stacey

doi:10.1016/j.robot.2003.11.002

Cited by 39 publications

(27 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Recent surveys show that neural networks account for the basis of approximately 40% of such controllers, while genetic programming is used in another 30% [10]. Progress has been made in the learning of simple behaviors such as locomotion and obstacle avoidance [12], phototaxis [16], and searching and foraging [13]. As research has moved to attempting tasks involving multiple behaviors, tension has developed between learning controllers for these complex tasks and limiting the amount of a priori background knowledge used in the learning frameworks.…”

Section: Introductionmentioning

confidence: 99%

Exploring and evolving process-oriented control for real and virtual fire fighting robots

Hardey

Corapcioglu

Mattis

et al. 2012

Proceedings of the 14th Annual Conference on Genetic and Evolutionary Computation

View full text Add to dashboard Cite

Current research in evolutionary robotics is largely focused on creating controllers by either evolving neural networks or refining genetic programs based on grammar trees. We propose the use of the dataflow languages for the construction of effective robotic controllers and the evolution of new controllers using genetic programming techniques. These languages have the advantages of being built on concurrent execution frameworks that lend themselves to formal verification along with being visualized as a dataflow graph. In this paper, we compare and contrast the development and subsequent evolution of one such process-oriented control algorithm. Our control software was built from composable, communicating processes executing in parallel, and we tested our solution in an annual fire-fighting robotics competition. Subsequently, we evolved new controllers in a virtual simulation of this parallel dataflow domain, and in doing so discovered and quantified more efficient solutions. This research demonstrates the effectiveness of using process networks as the basis for evolutionary robotics.

show abstract

Section: Introductionmentioning

confidence: 99%

Exploring and evolving process-oriented control for real and virtual fire fighting robots

Hardey

Corapcioglu

Mattis

et al. 2012

Proceedings of the 14th Annual Conference on Genetic and Evolutionary Computation

View full text Add to dashboard Cite

show abstract

“…In essence, EANNs are a specific type of artificial neural network that use a different method for learning in addition to the standard ANN methods. While standard ANNs can adapt to dynamic environments, EANNs' combination of evolutionary and neural learning allows them to adapt more quickly (Yao, 1999) and take advantage of temporal information as well (Nelson, Grant, Galeotti, & Rhody, 2004). In this regard, EANNs can be considered generic adaptive systems, which means they can change their architectures and learning methods to suit the problem without human involvement.…”

Section: An Evolutionary Artificial Neural Network (Eann) Is a Union mentioning

confidence: 99%

“…The two most commonly used types of EANNs are WEAs (Floreano & Mondada, 1998;Lee, 2003;Miglino, Lund, & Nolfi, 1995;Mondada & Floreano, 1995) and TEAs (Nelson, Grant, Galeotti, & Rhody, 2004; McKerrow, Learning to Avoid Objects and Dock with a Mobile Robot, 1999;Xu, Van Brussel, Nuttin, & Moreas, 2003). These are also the most common types used for the evolution of autonomous agents.…”

Section: Artificial Neural Network Configurationsmentioning

confidence: 99%

“…There are also robotic systems implemented with far more complex onboard processing systems which could easily handle an adaptive ANN structure outside of a simulated environment. However, typically the topologies of robotic systems are developed in a simulated environment before deployment into non-virtual robotic systems (Nelson, Grant, Galeotti, & Rhody, 2004 …”

Section: Artificial Neural Network Configurationsmentioning

confidence: 99%

“…Some systems that are evolved in simulation are subsequently implemented in real-world hardware to prove that the resulting system is realistically functional (Floreano & Mondada, 1998;Miglino, Lund, & Nolfi, 1995;Mondada & Floreano, 1995;Nelson, Grant, Galeotti, & Rhody, 2004;Ward, Zelinsky, & McKerrow, Learning to Avoid Objects and Dock with a Mobile Robot, 1999;Xu, Van Brussel, Nuttin, & Moreas, 2003). It is effectively impossible for a virtual environment to simulate all of the subtle details and variations of a real-world environment.…”

Section: Simulated Environments and Real-world Environmentsmentioning

confidence: 99%

See 2 more Smart Citations

Evolving artificial neural network controllers for autonomous agents navigating dynamic environments.

Lucas¹

View full text Add to dashboard Cite

Fungi Use Efficient Algorithms for the Exploration of Microfluidic Networks

Hanson

Nicolau

Filipponi

et al. 2006

Small

View full text Add to dashboard Cite

Fungi, in particular, basidiomycetous fungi, are very successful in colonizing microconfined mazelike networks (for example, soil, wood, leaf litter, plant and animal tissues), a fact suggesting that they may be efficient solving agents of geometrical problems. We therefore evaluated the growth behavior and optimality of fungal space-searching algorithms in microfluidic mazes and networks. First, we found that fungal growth behavior was indeed strongly modulated by the geometry of microconfinement. Second, the fungus used a complex growth and space-searching strategy comprising two algorithmic subsets: 1) long-range directional memory of individual hyphae and 2) inducement of branching by physical obstruction. Third, stochastic simulations using experimentally measured parameters showed that this strategy maximizes both survival and biomass homogeneity in microconfined networks and produces optimal results only when both algorithms are synergistically used. This study suggests that even simple microorganisms have developed adequate strategies to solve nontrivial geometrical problems.

show abstract

Maze exploration behaviors using an integrated evolutionary robotics environment

Cited by 39 publications

References 16 publications

Exploring and evolving process-oriented control for real and virtual fire fighting robots

Exploring and evolving process-oriented control for real and virtual fire fighting robots

Evolving artificial neural network controllers for autonomous agents navigating dynamic environments.

Fungi Use Efficient Algorithms for the Exploration of Microfluidic Networks

Contact Info

Product

Resources

About