> Evolutionary intelligence & evolutionary robotics

We use simulated evolution to automatically design "artificial nervous systems" for robots.

Evolution has been one of the prominent processes to shape the animal brain; imitating this process by employing evolution-inspired algorithms to design "artificial nervous systems" therefore drew considerable attention in artificial intelligence over the last two decades. This line of thought drives us to propose many methods to exploit evolutionary algorithms (EA) to design neural networks (neuro-evolution). We especially work on:

Collaborations

Students

Funding

Funding provided by the ANR (projects Evoneuro and Creadapt).

(images: S. Magenat & R. Wheeler)

Encouraging behavioral diversity: making EAs faster and better

Evolutionary Robotics (ER) aims at automatically designing robots or controllers of robots without having to describe their inner workings. To reach this goal, ER researchers primarily employ phenotypes that can lead to an infinite number of robot behaviors and fitness functions that only reward the achievement of the task---and not how to achieve it. These choices make ER particularly prone to premature convergence.

To tackle this problem, several papers recently proposed to explicitly encourage the diversity of the robot's behaviors, rather than the diversity of the genotypes as in classic evolutionary optimization. Such an approach avoids the need to compute distances between structures and the pitfalls of the non-injectivity of the phenotype/behavior relation; however, it also introduces new questions: how to compare behavior? should this comparison be task-specific? and what is the best way to encourage diversity in this context?

In our work, we proposed several methods to encourage behavioral diversity, in particular using multi-objective evolutionary algorithms (multi-objectivization). We compared each approach on three different tasks and two different genotypes.Results show that fostering behavioral diversity substantially improves the evolutionary process in the investigated experiments, regardless of genotype or task. Among the benchmarked approaches, multi-objective methods were the most efficient and the generic, Hamming-based, behavioral distance was at least as efficient as task-specific behavioral metrics.




Publications

  • Mouret, J.-B. and Doncieux, S. (2012). Encouraging Behavioral Diversity in Evolutionary Robotics: an Empirical Study. Evolutionary Computation. Vol 20 No 1 Pages 91-133. [ PDF | BIB ]
  • Doncieux, S. and Mouret, J.-B. (2013) Behavioral Diversity with Multiple Behavioral Distances. Proc. of IEEE Congress on Evolutionary Computation, 2013 (CEC 2013). Pages 1-8. [pdf]
  • Mouret, J.-B. (2011). Novelty-based Multiobjectivization. New Horizons in Evolutionary Robotics: Extended Contributions from the 2009 EvoDeRob Workshop, Springer, publisher. Pages 139--154. [ PDF | BIB ]
  • Doncieux, S. and Mouret, J.-B. (2010). Behavioral diversity measures for Evolutionary Robotics. WCCI 2010 IEEE World Congress on Computational Intelligence, Congress on Evolutionary Computation (CEC). Pages 1303--1310. [ PDF | BIB ]
  • Mouret, J.-B. and Doncieux, S. (2009). Overcoming the bootstrap problem in evolutionary robotics using behavioral diversity. IEEE Congress on Evolutionary Computation, 2009 (CEC 2009). Pages 1161 - 1168. [ PDF | BIB ]
  • Mouret, J.-B. and Doncieux, S. (2009). Using Behavioral Exploration Objectives to Solve Deceptive Problems in Neuro-evolution. GECCO'09: Proceedings of the 11th annual conference on Genetic and evolutionary computation , ACM, publisher. Pages 627--634.
[top]

Crossing the reality gap: the transferability approach

How to make sure that results obtained in simulation work on the robot?
In robotics, gradient-free optimization algorithms (e.g. evolutionary algorithms) are often used only in simulation because they require the evaluation of many candidate solutions. Nevertheless, solutions obtained in simulation often do not work well on the real device. The transferability approach aims at crossing this gap between simulation and reality by automatically making the optimization algorithm aware of the limits of the simulation.

Publications

  • Koos, S.and Mouret, J.-B.and Doncieux, S.(2013). The Transferability Approach: Crossing the Reality Gap in Evolutionary Robotics.
    IEEE Transactions on Evolutionary Computation. Vol 17 No 1 Pages 122 - 145 . [ PDF| DOI| BIB]
  • Mouret, J-Band Koos, S.and Doncieux, S.(2012). Crossing the Reality Gap: a Short Introduction to the Transferability Approach.
    Proceedings of the workshop ''Evolution in Physical Systems''. ALIFE. [ PDF| BIB]
  • Pinville, T.and Koos, S.and Mouret, J-B.and Doncieux, S.(2011). How to Promote Generalisation in Evolutionary Robotics: the ProGAb Approach.
    GECCO'11: Proceedings of the 13th annual conference on Genetic and evolutionary computation ACM, publisher . Pages 259--266. [ PDF| BIB]
  • Koos, S.and Mouret, J-B.(2011). Online Discovery of Locomotion Modes for Wheel-Legged Hybrid Robots: a Transferability-based Approach.
    Proceedings of CLAWAR, World Scientific Publishing Co., publisher. Pages 70-77. [ PDF| BIB]
  • Koos, S.and Mouret, J.-B.and Doncieux, S.(2010). Crossing the Reality Gap in Evolutionary Robotics by Promoting Transferable Controllers.
    GECCO'10: Proceedings of the 12th annual conference on Genetic and evolutionary computation ACM, publisher . Pages 119--126. [ PDF| BIB]

[top]

Encoding large neural networks: generative and developmental systems

EvoNeuro encoding

Neuro-evolution and computational neuroscience are two scientific domains that produce surprisingly different artificial neural networks. Inspired by the "toolbox" used by neuroscientists to create their models, this work investigate two main points: (1) neural maps (spatially-organized identical neurons) should be the building blocks to evolve neural networks able to perform cognitive functions and (2) well-identified modules of the brain for which there exists computational neuroscience models provide well-defined benchmarks for neuro-evolution.

Relationships between synaptic plasticity and generative and developmental systems (GDS)

We analyzed the mutual relationships between generative and developmental systems (GDS) and sy-naptic plasticity when evolving plastic artificial neural networks (ANNs) in reward-based scenarios. We first introduce the concept of synaptic Transitive Learning Abilities (sTLA), which reflects how well an evolved plastic ANN can cope with learning scenarios not encountered during the evolution process. We subsequently report results of a set of experiments designed to check that (1) synaptic plasticity can help a GDS to fine-tune synaptic weights and (2) that with the investigated generative encoding (EvoNeuro), only a few learning scenarios are necessary to evolve a general learning system, which can adapt itself to reward-based scenarios not tested during the fitness evaluation.

Publications

  • Mouret, J.-B. and Doncieux, S. and Girard, B. (2010). Importing the Computational Neuroscience Toolbox into Neuro-Evolution---Application to Basal Ganglia.
    GECCO'10: Proceedings of the 12th annual conference on Genetic and evolutionary computation ACM, publisher . Pages 587--594.[ PS | PDF | BIB ]
  • Tonelli, P. and Mouret, J.-B. (2011). On the Relationships between Synaptic Plasticity and Generative Systems.
    Proceedings of the 13th Annual Conference on Genetic and Evolutionary Computation. Pages 1531--1538. (Best paper of the Generative and Developmental Systems (GDS) track).[ PS | PDF | DOI | BIB ]

Overview of the development process. From left to right: (1) the genotype is a labeled graph with evolvable labels; (2) the labels are interpreted to a neuroscience-inspired description of the neural network; (3) for a given size of maps, this neural network can be fully developed into a neural network (for instance to evaluate its fitness).


[top]