Abstract
Neuro-evolution is an appealing approach to generating collective behaviors for robot swarms. In its typical application, known as off-line automatic design, the neural networks controlling the robots are optimized in simulation. It is understood that the so-called reality gap, the unavoidable differences between simulation and reality, typically causes neural network to be less effective on real robots than what is predicted by simulation. In this paper, we present an empirical study on the extent to which the reality gap impacts the most popular and advanced neuro-evolutionary methods for the off-line design of robot swarms. The results show that the neural networks produced by the methods under analysis performed well in simulation, but not in real-robot experiments. Further, the ranking that could be observed in simulation between the methods eventually disappeared. We find compelling evidence that real-robot experiments are needed to reliably assess the performance of neuro-evolutionary methods and that the robustness to the reality gap is the main issue to be addressed to advance the application of neuro-evolution to robot swarms.
Highlights
Neuro-evolution is an appealing approach to generating collective behaviors for robot swarms
We present the results of an empirical study in which we assessed and compared some of the most advanced neuro-evolutionary methods for the off-line design of robot swarms
The methods comprised in the study include: (a) Covariance Matrix Adaptation Evolutionary Strategy (CMA-ES)[42], for generating both single- and multi-layer perceptrons
Summary
Neuro-evolution is an appealing approach to generating collective behaviors for robot swarms. The neuro-evolutionary approach appears to be appropriate in swarm robotics[5] because it bypasses the main problem that designers face: defining what the individual robots should do so that the desired collective behavior emerges from their interactions. Likewise other more or less related optimization-based design methods[23,24,25,26,27,28], bypasses the problem of explicitly reducing the desired collective behavior to the one of the individuals, it appears to be, together with the other optimization-based methods, the only truly general approach to realizing robot swarms. There is a general understanding that, due to the so-called reality gap36,37—that is, the unavoidable difference between simulation models and the real world—results obtained in simulation cannot be considered as a valid assessment of a neuro-evolutionary method for the automatic off-line design of robot swarms. Some recent results indicate that the reality gap is a relative problem with some design methods that are affected to a great extent, while others appear to be intrinsically more robust[40]
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
