Craig W. Reynolds. Competition, coevolution and the game of tag. In Rodney A. Brooks and Pattie Maes, editors, Proceedings of the Fourth International Workshop on the Synthesis and Simulation of Living Systems, pages 59--69, Cambridge, MA, USA, 1994. MIT Press.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....coordination of multiple manipulators, articulated arms, or multi fingered hands, etc. human robot cooperative systems, and userinterface issues that arise with multiple robot systems [184] 8] 124] 1] the competitive subclass of collective behavior, which includes pursuit evasion [139], 120] and one on one competitive games [12] Note that a cooperative team strategy for, e.g. work on the robot soccer league recently started in Japan [87] would lie within our present scope. emerging technologies such as nanotechnology [48] and Micro Electro Mechanical Systems [117] that ....

C. Reynolds. Competition, coevolution and the game of tag. In Proc. A-Life IV, 1994.

....coevolution is often that it provides a useful way of dealing with the problems associated with static fitness landscapes (Bullock 1995, section 5) It appears that few of those working with artificial selection intentionally use coevolution as a step towards intrinsic evolution. Notably, Reynolds (1994) of Boids fame worked towards more automatic evolution by coevolving simulated mobile agent controllers which competed with each other in games of Tag. This eliminated the need to design a controller in order to evolve a controller, as in his previous work (Reynolds 1992) mentioned above. 4.2 ....

Reynolds, C. W. (1994). Competition, coevolution and the game of Tag. In R. A. Brooks and P. Maes (Eds.), Proceedings of Artificial Life IV, pp. 59--69. Cambridge, MA: MIT Press.

....with predator prey interactions such as Lotka Volterra systems, or with host parasite interactions. On the other hand, over the last decade, coevolutionary scenarios have been studied through direct competition of two or more species in Artificial Life style computer simulations (e.g. [12, 15, 22]) The evolutionary patterns observed there are often very complex, and a clear theoretical understanding of these patterns has not been obtained yet. Finally, Class ii) which lies between classes i) and iii) and which could therefore yield many new insights about the connection between ....

Craig W. Reynolds. Competition, coevolution and the game of tag. In Rodney A. Brooks and Pattie Maes, editors, Artificial Life IV, pages 59--69. MIT Press, 1994.

....from the fact that for many machine learning domains a suitable objective metric of performance is simply not available. Examples where coevolution provides a metric of performance that would otherwise be unavailable include the coevolution of pursuit and evasion behaviors [Miller and Cliff 1994, Reynolds 1994], and competitive manipulation of physical objects [Sims 1994] Apart from this primary benefit of providing some target for performance, coevolution is commonly understood to have several other benefits. The following list uses examples from the domain of chess but the concepts apply equally to ....

Reynolds, CW, 1994, Competition, Coevolution and the Game of Tag, in the proceedings of Artificial Life IV, R. Brooks and P. Maes, Editors, MIT Press, Cambridge, Massachusetts, pp 59-69.

.... from a predictive strategy, as opposed to reactive approach: more successful pursuers try to anticipate where their opponent 2 will go, based on that opponent s current heading and intentions and the environmental constraints they face (obstacles and boundaries) and then try to cut them off [Reynolds, 1994]. Evasion: Animals move away from things that threaten them. Again, if the threatening object is inanimate, we have a degenerate case of obstacle avoidance, or one step evasion . If the threat is animate, however, and does not wish to be evaded, then it will pursue, and sustained evasion becomes ....

Reynolds, C. W. (1994). Competition, coevolution and the game of tag. In R. A. Brooks and P. Maes (Eds.), Artificial Life IV, pages 59--69. Cambridge: MIT Press.

....this problem, Bullock, 1995] proposes using diffuse coevolution: allowing each species to compete against multiple opposing species, which are all evolved separately. Since each agent is tested against a wider diversity of opponents, more general strategies can evolve. Similar ideas are echoed in [Reynolds, 1994]. The goal of this research was to compare true and diffuse coevolution using a relatively sophisticated competition. Our results support the hypothesis that diffuse coevolution produces better agents than true coevolution. We chose a predator prey scenario to study coevolution. While one of the ....

....to evolve. Hence there is a trend in predator prey coevolution towards richer environments. The environment used by Reynolds to evolve tag players was a simple one: agents were modeled as circles existing in a discrete time, flat, two dimensional world without momentum, friction or other objects [Reynolds, 1994]. Reynolds states that a world with more realistic physics could produce more interesting motion and a rich environment for future studies. More complex physics was MERL TR 98 11 January 2 used in [Cliff and Miller, 1995] and [Cliff and Miller, 1996] This environment, while two dimensional, was ....

Reynolds, C. W. (1994). Competition, coevolution and the game of tag. In Brooks, R. and Maes, P., editors, Proceedings of the Fourth Workshop on Artificial Life, pages 59--69, Boston, MA. MIT Press.

....continuous space pursuer evader domain. The purpose of replacing (stateless) minimax strategies with evolvable agent controllers is to place the pursuit and evasion roles within a coevolutionary setting such that an arms race between ever more sophisticated evasion and pursuit strategies arises [16, 19, 24, 6, 7, 12, 26]. Unfortunately, the envisaged coevolutionary arms race towards complexity is yet to be substantially realized, or, at least, observed; a number of problematic issues surrounding coevolutionary techniques are known to exist, such as the Red Queen Effect [6] mediocre stablestates [22] and the ....

C. Reynolds. Competition, coevolution and the game of tag. In R. A. Brooks and P. Maes, editors, Artificial Life IV, pages 59--69. MIT Press, 1994.

....and explored to overcome this problem. Where the method described above required every program to compete in a match against every other program per generation, a strategy proposed by Reynolds only requires every program to compete against a small number of other programs in each generation [Rey94a] This new versus several strategy was tested in a domain where programs competed in games of Tag. Programs competed in seven games against different opponents, with the fitness of a program averaged across the number of games it played. This method reduces the number of games played from ....

....possible combinations in genetic programming is referred to as the state search space [Mon94] By enforcing typing restrictions on the placement of nodes as arguments to functions, the number of possible combinations in the state search space is dramatically reduced. In the Game of Tag problem [Rey94a] Reynolds also explores ways to reduce the state search space by placing restrictions on the composition of trees. In one of the described experiments, genetic programs were constructed with two explicit trees designed for different tasks one which was executed when the program was not it, ....

Craig Reynolds. Competition, coevolution and the game of tag. In Proceedings of the Fourth International Workshop on the Synthesis and Simulation of Living Systems, pages 59--69, July 1994.

....each other, usually a move is determined for each, and then executed, implying that each move takes the same length of time to execute. GPRobots differs from this, in that all instructions take an argument, e.g. how far to move forward, how many degrees to turn left etc. We share the view of [Reynolds, 1994a] that more complex or longer moves should take more than one time unit to execute. A relatively new programming paradigm is that of event driven programming. Event driven programming differs from most paradigms in that instead of writing a sequential program, a programmer writes functions to ....

....this sort of measurement to some degree. Only to some degree because, although it is easy to compare k individuals, where k is the number involved in a tournament, it is another matter to compare N individuals, where N is the size of the population. A variety of these methods were reported on by [Reynolds, 1994a] and are summarized in table 7.2. This paper uses a method which employs both the New versus several [Reynolds, 1994a] and the New versus neighbour [D Haeseleer, 1994] Individuals live in a one dimensional neighbourhood, and are chosen in groups of k for tournaments. Individuals in this ....

[Article contains additional citation context not shown here]

Reynolds, C. (1994a). Competition, coevolution and the game of tag. In ALife IV.

.... ] approach where the population must continually improve itself) A dynamic fitness function could be pre defined but dynamic GP fitness functions are often produced by co evolution [ Hillis, 1992; Angeline and Pollack, 1993; Angeline, 1993; Angeline and Pollack, 1994; Koza, 1991; Jannink, 1994; Reynolds, 1994; Ryan, 1995 ] Where it is felt certain characters will be required in the problem s solution the initial population and crossover can be controlled in order to ensure individuals within the population have these properties ( Langdon, 1995 ] and [ Langdon, 1996b ] have described ways in which ....

....are exploiting to achieve high fitness on the test case but at the expense of not generalising to the problem as a whole. Co evolution can provide an automatic means of dynamically changing the fitness function [ Siegel, 1994 ] There is increasing interest in using co evolution [ Sen, 1996; Reynolds, 1994; Ryan, 1995 ] and improved performance has been claimed [ Hillis, 1992 ] However a more dynamic framework makes analysis of population behaviour harder. In GP runs the concentration of primitives and variety within the population should be monitored (both can be done with little overhead) ....

Craig W. Reynolds. Competition, coevolution and the game of tag. In Rodney A. Brooks and Pattie Maes, editors, Proceedings of the Fourth International Workshop on the Synthesis and Simulation of Living Systems, pages 59--69, MIT, Cambridge, MA, USA, 6-8 July 1994. MIT Press.

....1 This is a characterizationof the definition itself; the leadingmotivationin [Mat94a] was actually the social nature of intelligence and its manifestation in group behaviors. 2 In this work, we do not discuss the competitive subclass of collective behavior, which includes pursuit evasion [Rey94, MC94] and one on one competitive games [AUN 94] Note that a cooperative team strategy for, e.g. the robot soccer league recently proposed by [Kit94] would lie within our present scope. 3 Even with this restriction, we find that in 7 years ( 1987 1994 ) well over 200 papers havebeen ....

C. Reynolds. Competition, coevolution and the game of tag. In Proc. A-Life IV, 1994.

....al. 14] have studied co evolution of machines and humans competing over the Internet and reported strategy improvement in both populations over time. In the context of adaptive autonomous agents, Koza [20, 21] applied Genetic Programming to the co evolution of pursuer evader behaviors, Reynolds [30] observed in a similar scenario that co evolving populations of pursuers and evaders display increasingly better strategies, and Sims used competitive coevolution to develop his celebrated artificial creatures [33] Cliff and Miller realised the potentiality of co evolution of pursuit evasion ....

....= 160 bits long while that of the prey was 5 times[20 synapses 2 thresholds] 110 bits long. Two separate populations of N individuals each were co evolved for g generations. Each individual was tested against the best competitors from k previous generations (a similar procedure was used in [33, 30, 2]) in order to improve co evolutionary stability. At generation 0, competitors were randomly chosen from the initial population, and later on they were randomly chosen from the pool of best individuals from previous generations (2 at the 3rd generation, 3 at 4th generation, 49 at 50th ....

C. W. Reynolds. Competition, Coevolution and the Game of Tag. In R. Brooks and P. Maes, editors, Proceedings of the Fourth Workshop on Artificial Life, pages 59--69, Boston, MA, 1994. MIT Press.

....for the Tic Tac Toe game. More recently, Rosin and Belew (1997) compared various co evolutionary strategies for discovering robust solutions to complex games. In the context of adaptive autonomous agents, Koza (1991, 1992) applied Genetic Programming to the coevolution of pursuer evader behaviors, Reynolds (1994) observed in a similar scenario that co evolving populations of pursuers and evaders display increasingly better strategies, and Sims used competitive co evolution to develop his celebrated artificial creatures (Sims, 1994) Cliff and Miller realised the potentiality of co evolution of ....

....bits long while that of the prey was 5 x (20 synapses 2 thresholds) bits long. Two separate populations of N individuals each were co evolved for g generations. Each individual was tested against the best competitors from k previous generations (a similar procedure was used in (Sims, 1994; Reynolds, 1994; Cliff and Miller, 1995) in order to improve co evolutionary stability. At generation 0, competitors were randomly chosen from the initial population, whereas in the remaining k Gamma 1 initial generations they were randomly chosen from the pool of available best individuals (2 at generation 3, ....

Reynolds, C. W. (1994). Competition, Coevolution and the Game of Tag. In Brooks, R. and Maes, P., editors, Proceedings of the Fourth Workshop on Artificial Life, pages 59--69, Boston, MA. MIT Press.

....there exist many versions of the PE game. The earliest formulations (Isaacs, 1965) are games of perfect information where optimal pursuit and evasion strategies follow analytically from agent capabilities. More recent work casts PE into a purely discrete, non kinematic reactive game (Koza, 1992; Reynolds, 1994), or into a continuous game of, essentially, imperfect information that incorporates a two dimensional physical model (Cliff and Miller, 1996) for example. Second, these differences in game formulation impact the nature of the solution space; where PE formulations have optimal solutions, the game ....

Reynolds, C. (1994). Competition, coevolution and the game of tag. In (Brooks and Maes, 1994), pages 59--69.

.... a set of techniques for analyzing and assessing adaptive progress of both populations [1] Artificial co evolution of competitive species has been studied also by other researchers using similar methods, such as Ray s Tierra system [13] Sim s creatures [15] and Reynolds pursuer evader sytems [14]. In very recent work, which will be briefly summarized 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 ....

.... 2 thresholds) bits long while that of the prey was 5 x (20 synapses 2 thresholds) bits long. Two populations of 100 individuals each were co evolved for 100 generations. Each individual was tested against the best competitors of the ten previous generations (a similar procedure was used in [15, 14, 1]) in order to improve co evolutionary stability. For each competition, the prey and predator were always positioned on a horizontal line in the middle of the environment at a distance corresponding to half the environment width (Figure 2, right) but always at a new random orientation. The ....

C. W. Reynolds. Competition, Coevolution and the Game of Tag. In R. Brooks and P. Maes, editors, Proceedings of the Fourth Workshop on Artificial Life, pages 59--69, Boston, MA, 1994. MIT Press.

....has failed in most other cases. Real time, interactive games (e.g. video games) have distinctive features that differentiate them from the better known board games. Koza [11 ch. 12] and others [17] evolved players for the game of Pacman. There has been important research in pursuer evader games [16, 13] as well as contests in simulated physics environments [20] But these games do not have human participants, as their environments are either provided by the game itself, or emerge from coevolutionary interactions inside a population of agents. 1.6 Coevolution of Interactive Adaptive Software In ....

Reynolds, C.W. (1994). Competition, Coevolution and the Game of Tag", Proceedings of Artificial Life IV. R. Brooks and P. Maes, eds. MIT Press.

....While the area of co evolution is growing, it is often overlooked and under appreciated. In Michalewicz s 1994 paper A Perspective on Evolutionary Computation [43] coevolutionary techniques get one paragraph and only one reference. Reynolds paper Competition, Co evolution and the Game of Tag [56] although not entirely relevant to this discussion because it uses Genetic Programming, contains one of the few surveys of related work. To the best of the author s knowledge it is, outside of this thesis, the largest survey on co evolution in GAs. During the review of relevant literature only ....

C. W. Reynolds. Competition, Coevolution and the Game of Tag. In R. Brooks and P. Maes, editors, Artificial Life IV, pages 59--69, 1994.

....In this work we examine the rise of cooperation strategies without implicit communication. This is achieved by having each predator agent being controlled by its own program. Such a system solves a cooperative co evolution problem as opposed to a competitive co evolution problem as described in [1, 7, 14]. We believe that cooperative co evolution provides opportunities to produce solutions to problems that can not be solved with implicit communication. Experimental Setup In our experiments, the initial configuration consisted of the prey in the center of a 30 by 30 grid, and the predators are ....

Craig W. Reynolds. Competition, coevolution and the game of tag. In Artificial Life IV. MIT Press, 1994.

....hands, etc. ffl human robot cooperative systems, and user interface issues that arise with multiple robot systems [Yokota et al. 1994] Arkin and Ali, 1994] Noreils and Recherche, 1991] Adams et al. 1995] ffl the competitive subclass of collective behavior, which includes pursuit evasion [Reynolds, 1994, Miller and Cliff, 1994] and one on one competitive games [Asada et al. 1994] Note that a cooperative team strategy for, e.g. work on the robot soccer league recently started in Japan [Kitano, 1994] would lie within our present scope. ffl emerging technologies such as nanotechnology [Drexler, ....

C. Reynolds. Competition, coevolution and the game of tag. In Proc. A-Life IV, 1994.

....find all (or most) near global optima. If the GA misses some peaks, then the GA s expertise will generalise poorly [5] A co evolutionary GA is one where each in dividual is evaluated by how well it performs against every other individual This is a promising way to learn game strategies [1] 5] [21] [22] 25] Without speciation, a GA will converge to only one high fitness solution, due to genetic drift caused by a small population. A coevolutionary GA will converge and overspecialise to only one high quality strategy for the game being learned. The overspecialised expertise thus created ....

Craig W. Reynolds. Competition, coevolution and the game of tag. In Artificial Life 4, pages 59--69. MIT Press, 1994.

.... the promising achievements described above, if one carefully looks at the results described in the literature focusing on competitive co evolution of pursuit evasion behaviors, it is easy to notice that coevolutionary benefits often come at the cost of several thousand individuals per population [26], several hundred generations [5] or repeated trials of evolutionary runs with alternating success [28] Moreover, since all the experiments have been conducted in simulation, often the results cannot be directly applied to real robots, either because agent descriptions are too abstract or ....

....described in section 2, but the predator had 5 more input units corresponding to the visual module. Two populations of 100 individuals each were co evolved for 100 generations. Each individual was tested against the best competitors of the ten previous generations (a similar procedure was used in [28, 26, 4]) in order to improve co evolutionary stability. For each competition, the prey and predator were always positioned on a horizontal line in the middle of the environment at a distance corresponding to half the environment width (Figure 6b) but always at a new random orientation. The competition ....

C. W. Reynolds. Competition, Coevolution and the Game of Tag. In R. Brooks and P. Maes, editors, Proceedings of the Fourth Workshop on Artificial Life, pages 59--69, Boston, MA, 1994. MIT Press.

....Toe game. More recently, Rosin and Belew (1997) compared various co evolutionary strategies for discovering robust solutions to complex games. In the context of adaptive autonomous agents, Koza [21, 22] applied Genetic Programming to the co evolution of pursuer evader behaviors, Reynolds (1994) [30] observed in a similar scenario that co evolving populations of pursuers and evaders display increasingly better strategies, and Sims used competitive co evolution to develop his celebrated artificial creatures [32] Cliff and Miller realised the potentiality of co evolution of pursuit evasion ....

C. W. Reynolds. Competition, Coevolution and the Game of Tag. In R. Brooks and P. Maes, editors, Proceedings of the Fourth Workshop on Artificial Life, pages 59--69, Boston, MA, 1994. MIT Press. Evolutionary Robotics in Behavior Engineering and Artificial Life 27

....basic premise of co evolution is that if one population devises a good ploy, then the other population will construct a counter to that ploy. We expect that the populations will see saw between being better on the average. This has been shown in Reynold s work on co evolution in the game of tag [25]. In his work, the two opposing agents, from the same population, take turns being 2 3 1 4 P Figure 5: Deadlock scenario for STGP but not MD. the predator and the prey. Whereas in our work, there are separate populations and the predator population has to manage cooperation between multiple ....

....We are examining the rise of cooperation strategies without implicit communication [9] This is achieved by having each predator agent being controlled by its own program. Such a system solves a cooperative co evolution problem as opposed to a competitive co evolution problem as described in [2, 8, 25]. We believe that cooperative co evolution provides opportunities to produce solutions to problems that can not be solved with implicit communication. 7 Conclusions Strongly typed genetic programming is able to generate effective individual behavioral strategies for predator agents trying to ....

Craig W. Reynolds. Competition, coevolution and the game of tag. In Artificial Life IV. MIT Press, 1994.

....The basic premise of coevolution is that if one population devises a good ploy then the other population will construct a counter to that ploy. We expect that the populations will see saw between being better on the average. This has been shown in Reynold s work on coevolution in the game of tag [14]. In his work, the two opposing agents, from the same population, take turns being the predator and the prey. Whereas in our work, there are separate populations and the predator population has to manage cooperation between multiple agents. What this research is going to show is that a prey ....

Craig W. Reynolds. Competition, coevolution and the game of tag. In Artificial Life IV. MIT Press, 1994.

....strategy is not required. Angeline and Pollack [AP93] used absolute and competitive fitness to evolve players for the game of Tic Tac Toe (TTT) They found that co eveolved players developed better strategies than players using absolute fitness measures. Similar results were obtained by Reynolds [Rey94a] when he used competitive fitness for the evolution of vehicle steering control programs used for the game of tag. He reports that near optimal solutions were found solely from direct competition between individuals in the population. 4.5.2 Multi Phase Fitness Andre [And94b] used GP to evolve ....

Craig W. Reynolds. Competition, coevolution and the game of tag. In Rodney A. Brooks and Pattie Maes, editors, Proceedings of the Fourth Internationa Workshop on the Synthesis and Simulation of Living Systems, pages 59--69. MIT Press, 1994.

....by making changes to the environment rather than explicitly by exchanging messages with each other. Koza used GP to evolve reactive agents for the central food foraging problem [Koz92] Haynes evolved agents for the predator prey (capture) problem [HWSS95] and others [HW95] HSSW95] Reynolds [Rey94] co evolved agents to simulate the game of tag. Manela [Man93] employed GAs to optimise architectural features of agents (including simple communication) for the pursuit problem. In this paper we concentrate on using GP to evolve agents that communicate directly with each other, via message ....

....pointing north. The steering mechanism allows the orientation of the vehicles to be changed to any one of the 8 compass directions. The goal is to evolve the code for agents A and B such that they steer the vehicles to enable them to meet. The vehicle model closely follows that used by Reynolds [Rey94]. The vehicles are given energy proportional to the euclidian distance between them. The latter is an attempt to prevent the meet you at (x,y) strategy in which the agents obviate any need for communication having previously agreed to always move to the fixed position (x,y) A single energy unit ....

Craig W. Reynolds. Competition, coevolution and the game of tag. In Rodney A. Brooks and Pattie Maes, editors, Proceedings of the Fourth International Workshop on the Synthesis and Simulation of Living Systems, pages 59--69, Cambridge, MA, USA, 1994. MIT Press.

....always be appropriate for the other. The subfield of co evolutionary machine learning attempts to achieve just this, building on earlier work in genetic algorithms and genetic programming. Some pertinent applications include Prisoner s Dilemma (Axelrod, 1984; Haynes et al. 1995) the game of tag (Reynolds, 1994), tic tac toe (Angeline Pollack, 1993) and backgammon (Pollack et al. 1996) How does the notion of co evolutionary learning apply to human learning, and, in particular, fit into today s framework of intelligent tutoring The current trend in educational culture transfers control away from the ....

Reynolds, C. W. (1994). Competition, coevolution and the game of tag. In Brooks, R. A. & Maes, P.

....distance from the pursuer. 2 In order to apply incremental evolution, we generated pairs (G 0 ; G 1 ) by choosing different relative speeds of the evader with respect to the pursuer. Clearly, all else 1 Koza [8] evolved both pursuers and evaders using genetic programming. Recently, Reynolds [13] has used coevolution to evolve pursuers and evaders, and the merits of this task as a testbed for the evolution of adaptive behavior have been discussed in [11] 2 To be specific: the pursuer moves a distance of 1.0 in every time step, and there are a total of 50 time steps. The initial vector ....

C. Reynolds. Competition, coevolution and the game of tag. In Artificial Life IV, 1994.

....ffl coordination of multiple manipulators, articulated arms, or multi fingered hands, etc. ffl human robot cooperative systems, and userinterface issues that arise with multiple robot systems [184] 8] 124] 1] ffl the competitive subclass of collective behavior, which includes pursuit evasion [139], 120] and one on one competitive games [12] Note that a cooperative team strategy for, e.g. work on the robot soccer league recently started in Japan [87] would lie within our present scope. ffl emerging technologies such as nanotechnology [48] and Micro Electro Mechanical Systems [117] that ....

C. Reynolds. Competition, coevolution and the game of tag. In Proc. A-Life IV, 1994.

.... a set of techniques for analyzing and assessing adaptive progress of both populations [1] Artificial coevolution of competitive species has been studied also by other researchers using similar methods, such as Ray s Tierra system [13] Sim s creatures [15] and Reynolds pursuer evader systems [14]. In very recent work, which will be briefly summarized 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 ....

.... 2 thresholds) bits long while that of the prey was 5 x (20 synapses 2 thresholds) bits long. Two populations of 100 individuals each were co evolved for 100 generations. Each individual was tested against the best competitors of the ten previous generations (a similar procedure was used in [15, 14, 1]) in order to improve co evolutionary stability. For each competition, the prey and predator were always posiBits for one synapse Condition 1 2 3 4 5 1 sign strength 2 sign strength noise 3 sign Hebb rule rate Table 1: Genetic encoding of synaptic parameters for each co evolutionary condition. 1: ....

C. W. Reynolds. Competition, Coevolution and the Game of Tag. In R. Brooks and P. Maes, editors, Proceedings of the Fourth Workshop on Artificial Life, pages 59--69, Boston, MA, 1994. MIT Press.

....of players for the Tic Tac Toe game. Koza 1 The Red Queen is a figure, invented by novelist Lewis Carroll, who was always running without making any advancementbecause the landscape was moving with her. 1991, 1992) applied Genetic Programming to the evolution of pursuer evader behaviors and Reynolds (1994) observed in a similar scenario that co evolving populations of pursuers and evaders display increasingly better strategies. Cliff and Miller realised the potentiality of co evolution of pursuit evasion tactics in evolutionary robotics. In the first of a series of papers (Miller and Cliff, 1994) ....

.... the promising achievements described above, if one carefully looks at the results described in the literature focusing on competitive co evolution of pursuit evasion behaviors, it is easy to notice that coevolutionary benefits often come at the cost of several thousand individuals per population (Reynolds, 1994), several hundred generations (Cliff and Miller, 1996) or repeated trials of evolutionary runs with alternating success (Sims, 1994) Moreover, since all the experiments have been conducted in simulation, often the results cannot be directly applied to real robots, either because agent ....

[Article contains additional citation context not shown here]

Reynolds, C. W. 1994. Competition, Coevolution and the Game of Tag. In Brooks, R. and Maes, P., editors, Proceedings of the Fourth Workshop on Artificial Life, pages 59--69, Boston, MA. MIT Press.

No context found.

Craig W. Reynolds. Competition, coevolution and the game of tag. In Rodney A. Brooks and Pattie Maes, editors, Proceedings of the Fourth International Workshop on the Synthesis and Simulation of Living Systems, pages 59--69, Cambridge, MA, USA, 1994. MIT Press.

No context found.

Craig W. Reynolds. Competition, coevolution and the game of tag. In Rodney A. Brooks and Pattie Maes, editors, Proceedings of the Fourth International Workshop on the Synthesis and Simulation of Living Systems, pages 59--69, MIT, Cambridge, MA, USA, 6-8 July 1994. MIT Press.

No context found.

Craig W. Reynolds. Competition, coevolution and the game of tag. In Rodney A. Brooks and Pattie Maes, editors, Proceedings of the Fourth International Workshop on the Synthesis and Simulation of Living Systems, pages 59--69, Cambridge, MA, USA, 1994. MIT Press. Run 2 Test 6 Run 2 Test 8 Run 7 Test 0 Run 7 Test 8 Run 7 Test 12 Run 7 Test 17

No context found.

C. W. Reynolds. Competition, coevolution and the game of tag. In R. A. Brooks and P. Maes, editors, Proceedings of the Fourth International Workshop on the Synthesis and Simulation of Living Systems, pages 59--69, Cambridge, MA, 1994. The MIT Press.

No context found.

Craig W. Reynolds. Competition, coevolution and the game of tag. In Rodney Brooks and Pattie Maes, editors, Artificial Life IV: Proceedings of the Fourth International Workshop on the Synthesis and Simulation of Living Systems. MIT Press, 1994.

No context found.

Reynolds, C. W. (1994), Competition, Coevolution and the Game of Tag, Proceedings of the Fourth International Workshop on the Synthesis and Simulation of Living Systems, MIT Press, Cambridge, Massachusetts

No context found.

Craig W. Reynolds. Competition, coevolution and the game of tag. In Rodney A. Brooks and Pattie Maes, editors, Proceedings of the Fourth International Workshop on the Synthesis and Simulation of Living Systems, pages 59--69, MIT, Cambridge, MA, USA, 6-8 July 1994. MIT Press.

No context found.

Craig W. Reynolds. Competition, coevolution and the game of tag. In Rodney A. Brooks and Pattie Maes, editors, Arti cial Life 4, pages 59-69. MIT Press, 1994.

No context found.

C. W. Reynolds. Competition, coevolution and the game of tag. In Brooks and Maes [1], pages 59--69.

No context found.

C. W. Reynolds. Competition, coevolution and the game of tag. In R. A. Brooks and P. Maes, editors, Proceedings of Artificial Life IV, pages 59--69, Cambridge MA, 1994. MIT Press.

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