Evolutionary Robotics is a technique which has recently received growing attention from the robotics research community as it promises the automatic synthesis of controllers. Such an automated framework is based on the use of artificial evolution (optimisation) to reinforce the learning of robots populations, by effectively tuning the parameters of randomly generated sets of controllers. The controllers commonly used are artificial neural networks (ANNs).

Parallelisation of evolutionary algorithms has been extensively studied in the context of global optimisation, resulting in a significant speedup of the optimisation process. Besides, island-based models have proved that migration of individuals between independent runs of an algorithm improves the performance of the optimisation process, both in terms of function evaluations required and in terms of the quality of the solution obtained, providing a better balance between exploration and exploitation of the search space. This technique has also been applied to large dimensional and difficult engineering problems, as well as the training of Recurrent Artificial Neural Networks.

The island model paradigm seems very suitable for application to Evolutionary Robotics for various reasons: 

1) It could significantly speedup the design process by exploiting parallelism, while improving the quality of the solutions found.
2) It could relieve the experimenter from a significant part of the burden required to properly set up the evolutionary process. Cooperating algorithms via migration without well-tuned parameters have proved to work as well as single instances of a well-tuned algorithm in global optimisation.
3) When designing controllers for robots, the exchange of individuals corresponds to introducing different types of solutions in a pool of already existing ones. This might increase the diversity in a population and via the recombination of genetic material, it might endow agents with capabilities that would be extremely unlikely to evolve in one run. In other words, it could facilitate the progressive composition of a rich behavioural repertoire.

The main objective of this study is to perform the optimisation of a neuro-controller in an island model, with a vision to demonstrate empirically an improvement of the automatic design methodology.

Studies and Experiments 

Study 1: Single Agent Navigation

• Experiment 1a: Single Island/Environment Experiment
• Experiment 1b: Multiple Island/Environment Experiment

Study 2: Single Agent Robustness

• Experiment 2a: No Island Robustness
• Experiment 2b: Island Robustness
• Experiment 2c: Fault-Island Robustness

Study 3: Active vision & varying surface properties

• Experiment 3a: No Island Selection
• Experiment 3b: Island Selection 

Study 4: Docking problem

Parallelised Mars rover simulator

Our previous simulator has been modified in order to allow implementation of the island model. The picture below show the new architecture. The main controller is a GUI interface running paGMO libraries and dealing with all the stuff related to evolution. This controller creates islands of populations and then sends calls to a certain number of Mars rover simulators that evaluate a particular genotype and then send the fitness back to the controller. In this case, it is possible to run massively parallel simulation saving plenty of time as compared to the previous sequential approach. 

Researchers

Christos Ampatzis, Dario Izzo, Leopold Summerer, Martin Peniak, Barry Bentley, Angelo Cangelosi, Davide Marocco

Results

All the results are on the project wiki page, please click here to see them.  

Links

ARIADNET - Evolution in robotic islands

Project wiki 

Bibliography

C. Ampatzis, D. Izzo, M. Rucinski and F. Biscani (2009). ALife in the Galapagos: migration effects on neuro-controller design, Proceedings of the European Conference on Artificial Life (ECAL 2009) link.

M. Peniak, A. Cangelosi, D.Marocco (2008). Autonomous robot exploration of unknown terrain: A preliminary model of Mars Rover robot. 10th ESA Workshop on Advanced Space Technologies for Robotics and Automation. Noordwijk, November link.

M. Peniak, D.Marocco, A. Cangelosi (2009). Co-evolving controller and sensing abilities in a simulated Mars Rover explorer. IEEE Congress on Evolutionary Computation (CEC) 2009. Trondheim Norway, 18th-21nd May link.

M. Peniak, D. Marocco, S. Ramirez-Contla and A. Cangelosi (2009). An active vision system for navigating unknown environments: An evolutionary robotics approach for space research. In: Proceedings of IJCAI-09 Workshop on Artificial Intelligence in Space, Pasadena, California, 17-18 July 2009 link.