![Sample our Mathematics & Statistics journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5943/TOC-Subject-Sample-2021-Mathematics-Statistics.jpg"})
Abstract
This article is concerned with a fixed-size population of autonomous agents facing unknown, possibly changing, environments. The motivation is to design an embodied evolutionary algorithm that can cope with the implicit fitness function hidden in the environment so as to provide adaptation in the long run at the level of population. The proposed algorithm, termed mEDEA, is shown to be both efficient in unknown environments and robust to abrupt and unpredicted changes in the environment. The emergence of consensus towards specific behavioural strategies is examined, with a particular focus on algorithmic stability. Finally, a real-world implementation of the algorithm is described with a population of 20 real-world e-puck robots.
Acknowledgement
This work was made possible by the European Union FET Proactive Intiative: Pervasive Adaptation funding the Symbrion project under grant agreement 216342.
Notes
1. With the notable exception of works in the field of artificial life [Citation17–21]. However, that research area strongly differs in its core motivation to the work presented here as we ultimately aim towards engineering solutions.
2. 11 input neurons; 5 hidden neurons; 2 output neurons; 1 bias neuron. The bias neuron value is fixed to 1.0 and projects onto all hidden and output neurons.
3. Note that the simulation begins with each agent containing a randomly initialized genome.
4. Combining the regrow delay update scheme with the equation relating the required number of food items and population size leads to a quadratic equation with one positive solution. In this set-up, the optimal population size is strictly below 80 agents. However, this set-up is very aggressive as soon as 50+ agents are active.
5. Evolutionary Robotics Database: http://www.isir.fr/evorob
db/.
6. As a reminder, proximity sensor length is 64 (cf. snapshot in for interpretation).
7. The open-hardware design can be found at http://www.e-puck.org.
9. As a counter example, generation times do need to be equal across robots as longer generation times would lead to more opportunities for a genome to spread.
10. http://www.youtube.com/watch?v=_ilRGcJN2nA (Accessed 1 December 2010).
11. One robot was removed due to technical failure in a previous experiment.