Dario Floreano and Stefano Nol. God save the red queen! competition in coevolutionary robotics. In John R. Koza, Kalyanmoy Deb, Marco Dorigo, David B. Fogel, Max Garzon, Hitoshi Iba, and Rick L. Riolo, editors, Genetic Programming 1997: Proceedings of the Second Annual Conference, pages 398406, Stanford University, CA, USA, 13-16 1997. Morgan Kaufmann.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....oflPD with two adaptive players. The IPD literature for co adaptive players shows us several interesting behaviors. Each of these is an artifact of the red queen efikct, so called, because the red queen in Alice in Wonderland states that in her world you must keep running just to stand still [12]. In an analogous way, the performance of each player in the two sided learning problem is relative to that of its opponent. In other words, when one player adapts and the other uses a static strategy (as in our previous work) the performance of the adaptive player is absolute with respect to its ....

D. Floriano and S. Nolfi, S., God save the red queen!: Competition in co-evolutionary robotics, in: Proceedings of the Second International Conference on Genetic Programming, (1997) 398-406.

....in view of a certain behavior to be developed by the robot. However quite often the behavior obtained is radically different from expectation. Hence, the design of the fitness function is seldom straightforward and often critical. Co evolution has recently been discussed as an alternative, cf. [3], but is not general enough to cover typical situations like evolving robots in a given hostile environment. It is well known that learning in the life time of the individuals may accelerate evolution enormously. In fact we think that in vivo evolution of robots can not be realized without ....

D. Floreano and S. Nolfi. God save the red queen! competition in coevolutionary robotics. In J. R. Koza, K. Deb, M. Dorigo, D. B. Fogel, M. Garzon, H. Iba, and R. L. Riolo, editors, Genetic Programming 1997: Proceedings of the Second Annual Conference, Stanford University, CA, USA, 13-16 July 1997. Morgan Kaufmann.

....hand in view of a certain behavior to be developed by the robot. However quite often the behavior obtained is radically di erent from expectation. Hence, the design of the tness function is seldom straightforward and often critical. Co evolution has recently been discussed as an alternative, cf. [3], but is not general enough to cover typical situations like evolving robots in a given hostile environment. It is well known that learning in the life time of the individuals may accelerate evolution enormously. In fact we think that in vivo evolution of robots can not be realized without ....

D. Floreano and S. Nol. God save the red queen! competition in coevolutionary robotics. In J. R. Koza, K. Deb, M. Dorigo, D. B. Fogel, M. Garzon, H. Iba, and R. L. Riolo, editors, Genetic Programming 1997: Proceedings of the Second Annual Conference, Stanford University, CA, USA, 13-16 July 1997. Morgan Kaufmann.

....[27] Emmanuel, T. 177, 185] Enns, Russell, 168] 14 Genetic algorithms in robotics Erbudak, M. 37] Erkmen, A. M. 37] Espenschied, Kenneth S. 216] Fagg, Andrew H. 401] Falkenauer, Emanuel, 43, 44] Farritor, Shane, 263] Feng, S. 400] Fennel, Theron Randy, 16] Floreano, Dario, [110, 134, 181, 124, 264, 293] Fogarty, Terence C. 129, 180, 212] Fogel, David B. 349, 350, 351] Fogel, Lawrence J. 349, 350] Ford, Gary P. 45] Fraser, A. P. 68, 71, 88] Fu, Li Chen, 39] Fujimoto, Hideo, 21, 36] Fujimoto, S. 243] Fujimoto, Shinsaku, 155] Fukuda, T. 173, 191, 275, 294] Fukuda, Toshio, 69, ....

....Nagaya, E. 22] Nagy, R. N. 103] Naito, T. 250] Nakagama, Hayato, 109] Nakanishi, M. 436] Nakano, Kaoru, 109] Nakaoka, N. 120] Natowicz, Ren e, 409, 426, 427] Navon, Ronie, 309] Nearchou, A. C. 235] Nearchou, Andreas C. 307] Noguchi, N. 84] Nolfi, S. 110] Nolfi, Stefano, [147, 150, 184, 236, 293] Nordahl, Mats G. 241] Nordin, Peter, 198, 202, 204, 237, 238, 287, 322] Nose, and Matsuo, 200] Nose, Matsuo, 92, 115, 156, 187, 195, 247, 311] Odagiri, R. 250] Ogawa, Akio, 148] Oh, Kong Ping, 290] Ohsaka, Kazumasa, 155] Ohsaki, K. 243] Ohwi, J. 152] Okuma, Shigeru, 298] ....

[Article contains additional citation context not shown here]

Dario Floreano and Stefano Nolfi. God save the red queen! competition in co-evolutionary robotics. In Koza et al. [460], page ? y(conf.prog) ga97aFloreano.

....in view of the target behavior to be developed by the robot. However quite often the emerging behavior is radically different from expectation. In general, the design of the fitness function is seldom straightforward and often critical. Coevolution has recently been discussed as a way out, cf. [8, 7], but is not general and effective enough to cover the situations of interest for the present paper. On the other hand using reinforcement learning one faces the related problem that finding the right distribution of rewards is essential for success. The common problem is that in order to breed ....

D. Floreano and S. Nolfi. God save the red queen! competition in co-evolutionary robotics. In J. R. Koza, K. Deb, M. Dorigo, D. B. Fogel, M. Garzon, H. Iba, and R. L. Riolo, editors, Genetic Programming 1997: Proceedings of the Second Annual Conference, Stanford University, CA, USA, 13-16 July 1997. Morgan Kaufmann.

....the coevolution of strategies for two player pursuit and evasion games. Pursuit and evasion is omnipresent in nature, and an important concept in (evolutionary) robotics. Evolution of pursuit and evasion strategies has been studied both in simulations [Cliff Miller, 1996] and with real robots [Floreano Nolfi, 1997]. In cases where the evader and pursuer have identical properties, a good strategy for the evader is to move as fast as possible on a straight line away from the pursuer. To provide for more interesting dynamics, the pursuer and the evader can be given different properties by, for instance, making ....

Floreano, D. and S. Nolfi, God Save the Red Queen! Competition in CoEvolutionary Robotics, in J.R. Koza et al, eds., Genetic Programming 1997: Proceedings of the Second Annual Conference, pp. 398-406, (Morgan Kaufmann, 1997).

....The fitness of the prey is proportional to the time it manages to survive without being hit by the predator. The artificial brains of the two robots are artificial neural networks. plexity and eventually to behavioral cycles displaying rapid alternation of non trivial pursuit evasion strategies [3]. The most important concept is that the fitness function should leave space for free interaction between the robot and its environment; in other words, the fitness function should not be very detailed. This allows the robot to explore several different ways of solving a problem making evolution ....

D. Floreano and S. Nolfi. God Save the Red Queen! Competition in Co-evolutionary Robotics. In J. Koza, K. Deb, M. Dorigo, D. Fogel, M. Garzon, H. Iba, and R. L. Riolo, editors, Proceedings of the 2nd International Conference on Genetic Programming, San Mateo, CA, 1997. Morgan Kaufmann.

....of experiments were performed in simulation to understand the influence of various parameters, such as the number of tournaments with opponents from previous generations, crossover and mutation probabilities, replicability of the experiments, etc. A detailed analysis of these data is provided in [12]. Here we provide only a summary of the basic results and compare them to the results obtained with the real robots. Two populations of 100 individuals each were co evolved for 100 generations. Each individual was tested against the best opponents from the most recent 10 generations. Figure 7 ....

....which forces the predator to get closer to walls and collide if the prey is 4 Although individuals evolved in simulation do not behave in the same way when downloaded into the real robots. missed (right of figure 11) A description of additional behaviors obtained in simulations is given in [12]. 3.1 Machine Learning and natural adaptation The results described above indicate that co evolution between competing species with a relatively short generational overlap does not necessarily display the type of monotonic progress over time expected from the optimizationoriented approach that ....

D. Floreano and S. Nolfi. God Save the Red Queen! Competition in Co-evolutionary Robotics. In J. Koza, K. Deb, M. Dorigo, D. Fogel, M. Garzon, H. Iba, and R. L. Riolo, editors, Proceedings of the 2nd International Conference on Genetic Programming, San Mateo, CA, 1997. Morgan Kaufmann.

No context found.

Floreano, D. and Nolfi, S. (1997b). God Save the Red Queen! Competition in Co-evolutionary Robotics. In Koza, J., Deb, K., Dorigo, M., Fogel, D., Garzon, M., Iba, H., and Riolo, R. L., editors, Proceedings of the 2nd International Conference on Genetic Programming, San Mateo, CA.

.... below, we have investigated the potentiality of the Red Queen effect for evolutionary robotics, and showed that, with a suitable combination of realistic simulations and measuring techniques, competitive co evolution can develop a variety of efficient behaviors without any effort in fitness design [7]. However, none of these experimental researches systematically explored the role of ontogenetic adaptive behavior in co evolution of competing species. Although most of the evolved systems include some form of noise, it is difficult to say whether this plays an important role on the specific ....

.... the competitor s behavior In the attempt to investigate these issues in very simple settings, we have compared co evolution of competing species equipped with different types of simple adaptive controllers with results from previous experiments where the controllers were genetically determined [7]. 2 Method As often happens in nature, predators and preys belong to different species with different sensory and motor characteristics. Thus, we employed two Khepera robots, one of which (the Predator) was equipped with a vision module while the other (the Prey) had a maximum available speed ....

[Article contains additional citation context not shown here]

D. Floreano and S. Nolfi. God save the red queen! competition in co-evolutionary robotics. In J. Koza, K. Deb, M. Dorigo, D. Fogel, M. Garzon, H. Iba, and R. L. Riolo, editors, Proceedings of the 2nd International Conference on Genetic Programming, Stanford University, 1997.

....levels of complexity. 2. Co Evolving Predator and Prey Robots Several researchers have investigated co evolution in the context of predators and prey in simulation [1, 4, 5, 6] More recently, we have tried to investigate this framework first by using realistic simulations based on the Khepera [7, 8] and subsequently the real robots [9] In this section, we will first describe our experimental framework and the results obtained in a simple case. Then, we will describe other two experimental conditions more suitable to the emergence of arm races between the two competing populations. 2.1 ....

....A simulator developed and extensively tested on Khepera by some of us [13] was used. However some of the experiments described have also been successfully replicated on real [9] 1 The parameters used in the simulations described in this paper are mostly the same as in the simulation described in [7]. However, in these experiments we used a simpler fitness formula (a binary value instead of a continuous value proportional to the time necessary for the predator to catch the prey) Moreover, to keep the number of parameters as small as possible, we did not use crossover. In the previous ....

[Article contains additional citation context not shown here]

Floreano, D., Nolfi, S.: God Save the Red Queen! Competition in Co-Evolutionary Robotics. In Koza, J-R., Kalyanmoy, D., Dorigo, M., Fogel, D.B., Garzon, M., Iba, H., Riolo, R.L. (eds): Genetic Programming 1997: Proceedings of the Second Annual Conference, San Francisco, CA: Morgan Kaufmann (1997)

....the predator quickly turns in the attempt to visualise the prey which rotates around it, producing an entertaining dance. These experiments indicate that competitive co evolution is a promising technique for automatic gradual evolution of complex behaviours without effort in fitness design [11, 10]. In the following section we shall turn our attention to the study of systems that include more than two autonomous robots. 5. Issues in Collective Autonomous Robots Collective autonomous robotics deals with teams of several autonomous robots which are involved in a shared mission. Design and ....

D. Floreano and S. Nolfi. God save the red queen! competition in co-evolutionary robotics. In J. Koza, K. Deb, M. Dorigo, D. Fogel, M. Garzon, H. Iba, and R. L. Riolo, editors, Proceedings of the 2nd International Conference on Genetic Programming, San Mateo, CA, 1997. Morgan Kaufmann.

....while generating more complex behaviors than the previous experiment. In this section, I shall describe another way of reducing fitness design by augmenting the dynamical properties of the environment. Instead of having a single robot, two robots are co evolved in competition with each other [14]. Competitive co evolution has recently attracted considerable interest in the community of Artificial Life and Evolutionary Computation. In the simplest scenario of two co evolving populations, fitness progress is achieved at disadvantage of the other population s fitness. Although it is easy ....

D. Floreano and S. Nolfi. God save the red queen! competition in co-evolutionary robotics. In J. Koza, K. Deb, M. Dorigo, D. Fogel, M. Garzon, H. Iba, and R. L. Riolo, editors, Proceedings of the 2nd International Conference on Genetic Programming, Stanford University, 1997.

....competitors from the most recent 5 generations. Instantaneous fitness scores measured at each generation display oscillatory dynamics, as expected from the tightlyrelated dynamics and the Red Queen effect described above. A more detailed analysis of these data and of further analysis is given in [14, 15]. Here, I wish to point out that what these data tell us, especially the fitness scores of the best individuals, is that prey and predator rapidly (few that five generations) develop counter strategies to defeat the competitors. After only 15 generations, the two robots developed a variety of ....

D. Floreano and S. Nolfi. God Save the Red Queen! Competition in Coevolutionary Robotics. In J. Koza, K. Deb, M. Dorigo, D. Fogel, M. Garzon, H. Iba, and R. L. Riolo, editors, Proceedings of the 2nd International Conference on Genetic Programming, San Mateo, CA, 1997. Morgan Kaufmann.

.... below, we have investigated the potentiality of the Red Queen effect for evolutionary robotics, and showed that, with a suitable combination of realistic simulations and measuring techniques, competitive co evolution can develop a variety of efficient behaviors without effort in fitness design [7]. However, none of these experimental researches systematically explored the role of ontogenetic adaptive behavior in co evolution of competing species. Although most of the evolved systems include some form of noise, it is difficult to say whether this plays an important role on the specific ....

.... the competitor s behavior In the attempt to investigate these issues in very simple settings, we have compared co evolution of competing species equipped with different types of simple adaptive controllers with results from previous experiments where the controllers were genetically determined [7]. 2 Method As often happens in nature, predators and preys belong to different species with different sensory and motor characteristics. Thus, we employed two Khepera robots, one of which (the Predator) was equipped with a vision module while the other (the Prey) had a maximum available speed ....

[Article contains additional citation context not shown here]

D. Floreano and S. Nolfi. God save the red queen! competition in co-evolutionary robotics. In J. Koza, K. Deb, M. Dorigo, D. Fogel, M. Garzon, H. Iba, and R. L. Riolo, editors, Proceedings of the 2nd International Conference on Genetic Programming, Stanford University, 1997.

No context found.

Dario Floreano and Stefano Nol. God save the red queen! competition in coevolutionary robotics. In John R. Koza, Kalyanmoy Deb, Marco Dorigo, David B. Fogel, Max Garzon, Hitoshi Iba, and Rick L. Riolo, editors, Genetic Programming 1997: Proceedings of the Second Annual Conference, pages 398406, Stanford University, CA, USA, 13-16 1997. Morgan Kaufmann.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC