Brooks, R. (1992). Artificial life and real robots. In Varela, F. and Bourgine, P., editors, Toward a practice of autonomous systems: Proceedings of the First European Conference on Artificial Life, pages 3--10. MIT Press.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....two approaches: 1) using a real robot for the experiments, or (2) using a simulated robot. The pros and cons for each of the solutions cannot be listed here, since the literature would fill several book shelves. A short introduction into the discussion with additional references can be found in [4]. For our purposes, the real robot solution has been favored, for several reasons: i) a real robot comes along with real world problems, not with artificial ones. ii) The results of a machine learning technique like GP can impressively be demonstrated with a real walking robot, much more ....

R. Brooks. Artificial life and real robots. In European Conference on Artificial Life, pages 3 10, 1992.

....not of any help. But simulation is a good tool for finding inconsistencies . This means that the first condition could be a reason for using simulation. The question if it can also help in finding theories that can be used practically is a really di#cult one to answer. Rodney Brooks writes in [Brooks, 1992]: Previously we have been very careful to avoid using simulations for two fundamental reasons: 1. Without regular validation on real robots there is a great danger that much e#ort will go into solving problems that simply do not come up in the real world with a physical robot. 2. There is a ....

Rodney A. Brooks. Artificial life and real robots. In Francisco J. Varela and Paul Bourgine, editors, Towards a practice of autonomous systems: Proceedings of the first European Conference on Artificial Life (ECAL 91), pages 3--10. MIT Press, 1992.

....An alternative view of artificial life uses computation to control robots in a real physical world. Although in this approach the more fundamental philosophical issues are not as apparent, it has the enormous advantage in a practical sense of using the physical world at face value. As Brooks [6] understates the point: It is very hard to simulate the actual dynamics of the real world. My first answer to the spider s question is that we can only compare life to nonlife, that is, to the nonliving world from which life arises and evolves. Artificial life must be compared with a real or an ....

Brooks, R., 1992, Artificial life and real robots. In Toward a Practice of Autonomous Systems, F. Varela and P. Bourgine, eds, MIT Press, Cambridge, MA, pp. 3-10.

....[Brooks, 1991a, page 577] As Brooks points out, a complete model of the world is hard for a robot to construct because, The data delivered by sensors are not direct descriptions of the world as objects and their relationships [and] commands to actuators have very uncertain effects. Brooks, 1991b, page 5] Without ignoring the lessons of the early Seventies, the nascent area of Cognitive Robotics [Lesprance, et al. 1994] seeks to reinstate the ideals of the Shakey project, namely the construction of robots whose architecture is based on the idea of representing the world by sentences of ....

R.A.Brooks, Artificial Life and Real Robots, Proceedings of the First European Conference on Artificial Life (1991), pp 3--10.

....and so does what is really evolved: connections weights; weights and connections; weights, connections, and number of neurones, etc. Programs Several authors propose the use of extended versions of genetic programming (GP) Koza92] to evolve programs capable of controlling the robot. [Brooks92] suggests the use of GP with a high level behavioural language. Bhanzaf97] uses GP to evolve assembly code, which maps sensorial inputs into actuator actions. Rule Based Systems [Dorigo93] and [Grefenstette94] use several forms of classifier systems, or rule based systems, where the rules are ....

R. Brooks, "Artificial Life and Real Robots", Towards a Practice of Autonomous Systems: Proceedings of the First European Conference on Artificial Life, MIT Press, Cambridge, MA, 1992, pp. 3-10.

....of natural selection instead of more complex engineering techniques. Evolutionary approaches to agent synthesis can be divided in three main groups according with the representation used for the individuals: Neural networks [4, 8, 10] Rule based systems [6, 7, 9] Computer programs [1, 2]. The choice of the most appropriate controller architecture for autonomous agents is the center of an ongoing discussion [2, 15] which will probably never end. But from it we can conclude that the use of very specific representations has obvious disadvan tages in the fact that when the approach ....

.... in three main groups according with the representation used for the individuals: Neural networks [4, 8, 10] Rule based systems [6, 7, 9] Computer programs [1, 2] The choice of the most appropriate controller architecture for autonomous agents is the center of an ongoing discussion [2, 15] which will probably never end. But from it we can conclude that the use of very specific representations has obvious disadvan tages in the fact that when the approach fails to find a solution to a given problem, a lot of effort is needed to start over, using a new representation and ....

R. Brooks, "Artificial Life and Real Robots", Towards a Practice of Autonomous Systems: Proceedings of the First European Conference on Artificial Life, MIT Press, Cambridge, MA, 1992, pp. 3-10.

....adaptation to the environment. Individual adaptation, especially learning, has been a topic of various researches. Despite of the success made, for example reinforcement learning, it has not solved all problems. We need also the evolutionary approach. Brooks himself has acknowledged this in [1]. In this paper we propose a method where behavior is not defined explicitly. Instead, it emerges from the neural network level. Furthermore the neural network, its input (sensors) and its output (effectors) will each adapt to the environment in a dynamic process. Thus the behavior is a dynamic ....

R. Brooks, "Artificial life and real robots," in Varela and Bourgine [5], pp. 3--10.

....earlier demonstrated to function on a real robot programmed 54 in the subsumption architecture [70] The primitive building blocks of the behavior programs are in this case the sensory inputs and action parameter outputs, Boolean connectives, conditionals, and the subsumption primitives. Brooks [23] has however criticised these results, mostly because the primitive building blocks were well chosen (based on an analysis of a known solution) and simplifying assumptions were made concerning the Boolean nature of certain conditionals. Off line evolution creates a new problem, which is the gap ....

....Off line evolution creates a new problem, which is the gap between the virtual world of the simulator and the real world. Koza [57] uses a very simple virtual world. Cliff, Husbands and Harvey [26] use a much more sophisticated simulation to test out the fitness of a solution. But, as Brooks [23] points out, the gap between simulated and real world will always remain quite large. One possible way out is to use the real robot as soon as reasonable solutions have been discovered. An example application of this technique is discussed in [105] The application concerns the optimisation of ....

Brooks, R. (1992) Artificial life and real robots. In: F.J. Varela and P. Bourgine (eds.) Toward a Practice of Autonomous Systems. Proceedings of the First European Conference on Artificial Life. MIT Press/Bradford Books, Cambridge Ma. p. 3-10.

....world (Harvey 1992) Adaptation need not be on an individual scale alone, as accumulative change over generations can be thought of as adaptation on a longer timescale. But what class of physical system should be evolved in this fashion Why not some programming language, as advocated by Brooks (Brooks 1992) There are good grounds for thinking that a generalised form of connectionist network could be one very appropriate class. Let us start with three basic axioms: 1. The brain should be a physical system, occupying a physical volume with a finite number of input and output points on its ....

Rodney A. Brooks. Artificial life and real robots. In Proceedings of the First European Conference on Artificial Life. MIT Press/Bradford Books, Cambridge, MA, 1992.

....15] A primary means of gathering distal sensory information is by use of visual sensing, so we believe visually guided agents should be studied from as early a stage as possible. ffl While we could impose on our robot some visual sensors with fixed properties, we advocate (in common with Brooks [6]) the concurrent evolution of visual sensor morphology and the control networks: separating morphology from control is a measure which is difficult to justify from an evolutionary perspective, and potentially misleading. ffl For reasons of parsimony, studies of visually guided agents should ....

R. A. Brooks. Artificial life and real robots. In F. J. Varela and P. Bourgine, editors, Toward a Practice of Autonomous Systems: Proceedings of the First European Conference on Artificial Life (ECAL91), pages 3--10, Cambridge MA, 1992. M.I.T. Press --- Bradford Books.

....424] Bikdash, M. 296] Biondi, Joelle, 146] Blume, Christian, 102, 370] Bonarini, Andrea, 330] Boone, G. 86] Both, Hans Heinrich, 434] Boudreau, R. 259] Bradshaw, A. 103] Braunstingl, R. 127, 164] Bressgott, W. 33] Brevart, V. 128] Brillowski, K. 260] Brooks, Rodney A. [331] Browne, David, 211] Bruce, Wilker Shane, 289] Buckles, Bill P. 41] Bull, Lawrence, 129, 212] Bullock, G. N. 52] Burdick, Joel W. 170] Campbell, M. L. 100, 255] Chan, F. T. S. 30] Chan, K. K. 98] Chang, K. K. 431] Chatroux, Thierry, 332, 422, 423] Chedmail, P. 213, 261] ....

.... 150, 184, 236] biped, 234] control, 150, 157, 184, 236] hexapod, 216] manipulator, 251] manipulators, 135] manufacturing, 175] mobile, 347, 83, 147, 197, 258] path planning, 172] walking, 234] robot control, 408] robot programming, 396] robot societies, 137, 192] robotics, [327, 394, 410, 336, 337, 338, 339, 344, 350, 402, 432, 329, 331, 340, 326, 341, 342, 346, 347, 352, 353, 354, 355, 378, 379, 380, 395, 399, 401, 403, 405, 407, 416, 417, 59, 330, 332, 358, 360, 361, 362, 363, 364, 365, 366, 367, 373, 374, 381, 381, 382, 383, 384, 385, 386, 387, 388, 389, 392, 400, 406, 413, 415, 419, 420, 421, 422, 423, 431, 61, 68, 71, 72, 79, 90, 98, 101, 103, 106, 108, 109, 113, 114, 128, 145, 162, 169, 176, 205, 208, 229, 242, 246, 275, 281, 285, 287, 288, 293, 294, 299, 306, 309, 315, 316, 318] robotics animats, 144] autonomous, 390, 391, 77, 80, 86, 88, 110, 134, 148, 161, 180, 181, 232] autonomous agents, 151] biped, 206] cellular, 359, 69, 159, 173] collision avoidance, 65, 298] configuration, 126, 170] control, 78, 91, 117, 118, 136, 142, 167, 186, 190, 201, 218, 221, ....

[Article contains additional citation context not shown here]

Rodney A. Brooks. Artificial life and real robots. In Varela and Bourgine [462], pages 3--10. ga:Brooks91a.

.... explored at a hardware and firmware level [Lorigo97] Brooks has argued against the possibility of an adequate simulation of a physical system claiming that complex agent behavior is a reflection of a complex environment and that virtual reality is not rich enough to provide this level of stimulus [Brooks91]. Although this viewpoint is valid for the development of real world robots, if an agent s purpose is to ultimately function within a virtual environment, then the most appropriate place for its development is in virtual reality. This paper describes issues in using Java3D to develop a modified ....

Brooks, R., "Artificial Life and Real Robots", Towards a Practice of Autonomous Systems: European Conference on Artificial Life, MIT Press, Paris, France, pp. 3-10, 1991

....1996, p. 76) ffl As the complexity of robotic systems grows and the gap between the simulation and the real system widens, the question of the value of investing in a specialized simulation will become increasingly important. Mataric and Cliff, 1996, p. 76) Mataric and Cliff, in agreement with (Brooks, 1992), raise the concern that a lack of simulation fidelity can lead to problems of transference, where robotic controllers that evolve in simulation are unable to perform effectively when transferred to real robots because they exploit features of the simulator that are not found in the physical ....

Brooks, R. (1992). Artificial life and real robots.

....for building autonomous robots. Although many successful robots have been built based on this approach, increasing robot complexity makes the design difficult. Consequently, the evolutionary approach was advocated to provide some kind of design automation for building behavioral modules [3][4] The first work proposing to use the genetic approach to synthesize programs for robot control is [6] Due to being overly simplified, however, it has been criticized as being not complicated enough to control a real robot [3] Another example of using a GP approach to evolve control programs ....

.... some kind of design automation for building behavioral modules [3] 4] The first work proposing to use the genetic approach to synthesize programs for robot control is [6] Due to being overly simplified, however, it has been criticized as being not complicated enough to control a real robot [3]. Another example of using a GP approach to evolve control programs is given in a series of papers by Reynolds. In the final version [8] the author applied arithmetic operations, such as , Gamma, and the conditional operation iflte to calculate a single output value from sensor values ....

R. A. Brooks. Artificial Life and Real Robots. In Proceedings of the First European Conference on Artificial Life, p3-10, 1991.

....sensing and action and changes in the environment can require frequent replanning the cost of which may be prohibitive for complex systems. Planner based approaches have been criticized for scaling poorly with the complexity of the problem and consequently not allowing for reaction in real time (Brooks 1990b, Brooks 1991c) 2 The world may or may not be physical. Various attempts at achieving real time performance have been proposed. Perhaps the most prominent are purely reactive bottom up approaches which implement the agent s control strategy as a collection of preprogrammed condition action ....

....communication, one is used by the operating system, and one is used as the brain of the robot, for executing the down loaded control system used in the experiments. The control systems are programmed in the Behavior Language, a parallel programming language based on the Subsumption Architecture (Brooks 1990a) 1 The IRs are all the same frequency and mechanically positioned for obstacle detection rather than communication. 4.2.4 Hardware Limitations Properties of physical hardware impose restrictions not only on the control strategies that can be applied, but alson on the types of tasks and ....

[Article contains additional citation context not shown here]

Brooks, R. A. (1991a), Artificial Life and Real Robots, in `Toward A Practice of Autonomous Systems: Proceedings of the First European Conference on Artificial Life', The MIT Press.

....of autonomous sensorimotor agents date back to the end of the Eighties, 3 when all studies were still carried out in computer simulations. A few years later, the appearance of more robust, flexible, and user friendly robots, and a general awareness of the limitations of simulation methods [2], created a strong motivation for the first physical implementations of evolutionary robots [9, 17] Despite the importance of keeping in mind the hard constraints of operating with physical robots, simulations still play an important role in evolutionary robotics for at least two sets of ....

R. A. Brooks. Artificial Life and real robots. In F. J. Varela and P. Bourgine, editors, Toward a practice of autonomous systems: Proceedings of the First European Conference on Artificial Life. The MIT Press/Bradford Books, Cambridge, MA, 1992.

....aim is to apply fundamental principles of evolution and self organization to autonomous agents. The evolutionary approach applied to autonomous agents is today a rising research area. Various methods might be mentioned which make use of evolutionary processes on different structures. Rodney Brooks [3] proposed a high level language, GEN, which may evolve under genetic control, and which can be compiled into BL (Behavior Language) a lower level language, dedicated to a real mobile robot. He outlines the danger of the lack of realism of simulators but also acknowledges that such tools seem to ....

R.A. Brooks. Artificial life and real robots. In F. Varela and P. Bourgine, editors, Towards a Practice of Autonomous Systems, Proceedings of the First International Conference on Artificial Life, Paris. MIT Press, 1992.

....the behaviours designed to fulfill these tasks are well tailored to a particular situation. So far, no conceptual model of how behaviours are related to each other has been presented. Although some recent work has been reported on the use of genetic algorithm techniques within this approach [7], we do not expect this more engineering oriented 11 approach to behaviour based robotics to give any deeper insights in the evolutionary processes mentioned above. We have to look for other methodologies to explain the emergence of adaptive behaviour. Ethological models During the last 80 ....

R.A.Brooks, "Artificial Life and Real Robots," to appear in Proc. of the 1st European Conf. on Artificial Life, December 1991, Paris, France.

....danger of a combinatorial explosion, just as there is when a conventional planner searches its solution space. Of course workers with large rule based systems have already discovered that adding extra pieces of knowledge may have unforeseen results in a data driven system. Indeed Brooks now argues [10] that complex sets of behaviours should not be designed but be learned, using techniques such as genetic programming. It is not clear that any workers have produced a system more complex than the drinks can collector [9] and even this system is remarkably specific (it uses one corridor and one ....

Brooks, R. Artificial Life and Real Robots. Proceedings, 1st European Conference on Artificial Life, 1992

....structure rather than the more normal linear string. Crossover swaps complete sub trees, so always producing syntactically correct children. If the sub trees swapped are of different sizes then the child will (probably) not have the same genotype size as either parent. Rodney Brooks proposed, in [Brooks 1992], an extension of Koza s ideas with the aim of evolving robots. However, his Behavioural Language (BL) in effect forms blueprints rather than recipes for the robots networks. This is highly objectionable on grounds that will be given in part III. Fr d ric Gruau [1996] uses GP to evolve his ....

R.A. Brooks. Artificial life and real robots. In Proceedings of the First European Conference on Artificial Life. MIT Press/ Bradford Books, Cambridge, MA.

....becoming stuck due to repetition of an identical sensor motor sequence. 3 Reactive Controller May new ideas have been proposed to meet the challenges posed by mobile robots. Brooks decomposes a control system into a set of interacting behavior modules described in LISP like programming languages [2]. Others use production rules approach as building blocks for the control system [6] 9] or general dynamical systems [25] Some researchers propose to use artificial neural networks due to its generally smoother search space and its working with very low primitives avoiding using preconceptions ....

Rodney Brooks. Artificial life and real robots. In F. J. Varela and P. Bourgine, editors, Proceedings of the First European Conference on Artificial Life, pages 3--10, Cambridge, MA, 1992. MIT Press / Bradford Books.

.... the aim of which is to build autonomous system which can adapt to the world it is embedded in, this world being changing and often unpredictable [13] 20] To answer this new domain of robotics, Brooks has proposed a new designing methodology for mobile robots called subsumption architecture [2]. His approach consists to build behavior based robots. To deal with the autonomous arbitration of the multiple behaviors, Brooks outlines an evolutionary approach based on Koza s genetic programming techniques [12] He proposed a high level language, GEN, which could evolve, and then compiled ....

....obtain the desired behavior due to the small number of observable data obtained by the robot interaction with its environment. This generalizing ability of EHW is made effective through the exploitation by the EHW of the symmetries in the sensors motions mappings representing the desired behavior [2]. Because of the small input space, the reactive behavior can also be represented by the truth table form of the Boolean function implemented by a random access memory (RAM) In this case the sensor input is an address and its content is the action associated with this sensor input. We demonstrate ....

[Article contains additional citation context not shown here]

R. Brooks. Artificial life and real robots. In F. J. Varela and P. Bourgine, editors, Proceedings of the First European Conference on Artificial Life, pages 3--10, Cambridge, MA, 1992. MIT Press / Bradford Books.

....also stresses the importance of using a physical device (e.g. a robot) instead of a simulated agent. By using real robots, several additional factors due to the physical properties of the robot and of the environment must be taken into account (e.g. friction, inertia, ambient light, noise, etc. (Brooks, 1992). Moreover, only realistic types of sensors and actuators (instead of idealized ones that may not respect all the physical constraints or may have infinite precision) can be used. Similarly, the sensory inputs and the motor outputs should necessarily correspond to physical measures or forces (i.e. ....

Brooks, R.A. (1992) Artificial life and real robots, In F. J. Varela & P. Bourgine (Eds.), Toward a Practice of Autonomous Systems: Proceedings of the First European Conference on Artificial Life. Cambridge, MA: MIT Press/Bradford Books.

.... the aim of which is to build autonomous system which can adapt to the world it is embedded in, this world being changing and often unpredictable [20] 29] To answer this new domain of robotics, Brooks has proposed a new designing methodology for mobile robots called subsumption architecture [3]. His approach consists to build behavior based robots. To deal with the autonomous arbitration of the multiple behaviors, Brooks outlines an evolutionary approach based on Koza s genetic programming techniques [19] He proposed a high level language, GEN, which could evolve, and then compiled ....

....for mobile robots have been proposed. In the introduction we mention the subsumption approach which decomposes the control system into a set of interacting behavior modules. Each interacting behavior simulating a augmented finite state automaton, i.e. finite state machines with timing elements [3]. Other researchers in the domain of autonomous agent have proposed to use artificial neural Sensors Motors Left Wheel Right Wheel Proximity Direction Sensor Representation Motions Representation Reactive Robot Controller AND fuse array OR fuse array AND gate OR gate Bumping Distance ....

R. Brooks. Artificial life and real robots. In F. J. Varela and P. Bourgine, editors, Proceedings of the First European Conference on Artificial Life, pages 3--10, Cambridge, MA, 1992. MIT Press / Bradford Books.

....that have already been correctly translated from analog signals (e.g. sonar, infra red, etc. to symbols (e.g. door is open ) In response to such shortcomings, some researchers argue for the development of adaptive robots that evolve behaviors without using a pre specified model of the world [9]. Finally, there is a growing interest in the possibility of physical robots based on nano technology, that truly self replicate and evolve in adaptation to their environment [10] The genetic knowledge engineering paradigm treats the knowledge acquisition task for intelligent robots as a ....

Brooks, R. Artificial Life and Real Robots. Cambridge: MITPress. 1992.

....to execute adaptive control policies for unknown and changing environments (Williams Nayak 1996b) In the robotics domain, the design of such a component for simple adaptive control policies or behaviors has been proposed by Brooks. Brooks proposal is to combine a subsumption architecture (Brooks 1992) with an evolutionary system to build behavior based robots. Researchers have examined this new approach, with Beer, for example, exploring the evolution of a module controlled by continuous time recurrent neural networks for the locomotion control of a six legged insect like agent (Beer ....

.... In the introduction we mentioned the model based reactive approach which tries to merge a wide range of AI techniques (Williams Nayak 1996a) One of the technique which may be used is the subsumption approach which broke down the control system into a set of interacting behavior modules (Brooks 1992). Other researchers used artificial neural networks of some variety as the basic building blocks for the control system due to their generally smoother search space and their working with very low primitives avoiding using assumptions about the properties of the systems (Miglino, Lund, Nolfi ....

Brooks, R. 1992. Artificial life and real robots. In Varela, F. J., and Bourgine, P., eds., Proceedings of the First European Conference on Artificial Life, 3--10.

.... the best motion for each world state from a delayed reward signal [29] 33] For learning this simple robot task in unknown and dynamic environments, researchers have applied evolution based learning algorithms [22] to low level control architecture such as LISP like programming languages [19][4][26] finite state automata [32] production rules (classifier systems) 35] 6] 9] process network [28] and neural networks [7] 2] 24] 5] 10] 13] 12] In our experiments, the learning task consists to find the function F , mapping 256 inputs (world states) to 8 outputs (motions) in a search ....

Rodney Brooks. Artificial life and real robots. In F. J. Varela and P. Bourgine, editors, Proceedings of the First European Conference on Artificial Life, pages 3--10, Cambridge, MA, 1992. MIT Press / Bradford Books.

....modules for artificial nervous systems, or artificial embryology . Beer (Beer and Gallagher 1992) used GAs to synthesize a walking behaviour for a six legged agent. It is only recently that serious proposals have been made to use evolutionary approaches to real world robots (PRANCE 1991, Brooks 1992, Husbands and Harvey 1992) Most interest in such an approach has probably come from Japan, which currently has significantly more than half of all the robots in existence. Recently the Japanese government research laboratories, ATR in Kyoto, have set up a well funded research group for ....

....Robotics Group were the local organisers. Here what I consider to be the two most significant proposals (apart from that presented here) will be briefly summarised; that of Brooks, and that of Beer and Gallagher. Brooks evolutionary approach In December 1991 Brooks reported at ECAL 91 in Paris (Brooks 1992) that, following a suggestion by Langton, he was starting work on an evolutionary approach to robotics, based on Koza s Genetic Programming techniques (Koza 1990) He acknowledged the advantages of largely using simulations, but stressed the dangers of using simulated worlds rather than real ....

[Article contains additional citation context not shown here]

Rodney A. Brooks. Artificial life and real robots. In Proceedings of the First European Conference on Artificial Life. MIT Press/Bradford Books, Cambridge, MA, 1992.

....are easily avoided. As has been shown by several experimenters [20, 3, 23] it is possible to evolve controllers in simulation for a real robot. Now that this is no longer in doubt the question becomes one of whether the technique will scale up. In [22] and similar points were made earlier in [14, 6, 10]) the authors argue that if behavioural transference can only be guaranteed when a carefully constructed empirically validated simulation is used, then as robots and the behaviours we want to evolve for them become more complicated, so do the simulations. The level of complexity involved, they ....

R. Brooks. Artificial life and real robots. In F. J. Varela and P. Bourgine, editors, Toward a Practice of Autonomous Systems: Proceedings of the first European Conference on Artificial Life, pages 3--10, Cambridge, Massachusetts, 1992. MIT Press / Bradford Books.

....are less the product of the system design and more the result of adaptive behavior. 1.2 GP in the Real World Most previous experiments with GP and robot control have employed a simulated robot and environment. Judging from our own experience and from the general opinion in the research field (Brooks, 1992) we strongly believe in experiments with real robots for several reasons. Simulations often display several weaknesses when compared to experiments in the real world. The simulated world is often too ideal, resulting in the agent learning to react to small details in the simulation that will not ....

....a different approach for more complex problems. A useful strategy to create robust behavior is to divide complex actions into action primitives which only perform specialized tasks like avoiding obstacles or moving ahead, and then combine them again for the creation of higher levels of competence (Brooks, 1992). In our approach the GP system first learns the sub tasks and then evolves a higher level action selection strategy for deciding which of the evolved lower level algorithms should be in control. The lower level sub tasks are: ffl Obstacle avoidance. ffl Wall following. ffl Homing behavior. ....

Brooks, R. (1992). Artificial Life and Real Robots. In F.J. Varela and P.

....stage, transferring the end results to the real world. Since an evolutionary approach potentially requires the evaluation of populations of robots over many generations, a natural first thought is that simulations will speed up the process, making it more feasible. Despite initial scepticism (Brooks, 1992), it has recently been shown that control systems evolved in carefully constructed simulations, with an appropriate treatment of noise, transfer extremely well to reality, generating almost identical behaviors in the real robot (Jakobi, Husbands, Harvey, 1995; Thompson, 1995) However, both of ....

....in Viola (1988) De Garis (1992) proposed using GAs for building behavioral modules for artificial nervous systems, or artificial embryology. However, it is only recently that more complete propos 152 HUSBANDS et al. als have been made to use evolutionary approaches in robotics (Brooks, 1992; Husbands Harvey, 1992) Brooks (1992) outlines an approach based on Koza s genetic programming techniques (Koza, 1992) He acknowledged that time constraints would probably necessitate the use of simulations. However, he stressed the dangers of using simulated worlds rather than real worlds. ....

[Article contains additional citation context not shown here]

Brooks, R. A. 1992. Artificial life and real robots. In F. J. Varela & P. Bourgine (Eds.), Proceedings of the First European Conference on Artificial Life. Cambridge, MA: MIT Press/Bradford Books. Pp. 3--10.

....controllers. Unfortunately, a lack of fidelity in the simulator can lead to problems of transference; that is, controllers evolved in simulation do not account for the subtleties in the physical characteristics of the robots or the task environment and fail when transferred to real robots [Brooks, 1992, Mataric and Cliff, 1996] Transference problems can provably be eliminated through careful design of the simulator [Jakobi, 1997a, Jakobi, 1997b] but only by the assumption that the environmental factors critical for the task are known. Distributed robotics applications are particularly ....

....sensors and four contact points that collect power from the floor. 3.2 Continuous Power Technology The power requirements for embodied evolution demand a novel power delivery system. Battery power is able to sustain a robot only for a period on the order of hours, often no more than two or three [Brooks, 1992]. Longer periods of uninterrupted power can be achieved by either tethering a robot directly to a power source [Mondada and Floreano, 1996] or by providing battery recharge stations for the robot to visit periodically. Nevertheless, tethers easily tangle with only a few robots, and recharge ....

Brooks, R. (1992). Artificial life and real robots. In Varela, F. and Bourgine, P., editors, Proceedings of the First European Conference on Artificial Life, pages 3--10. MIT Press.

....Foundation Graduate Fellowship. Any opinions, findings, conclusions or recommendations expressed in this publication are those of the author and do not necessarily reflect the views of the National Science Foundation. on problems they claim will be integratable into a general framework. In [5], Brooks adds that cellular representations of space are problematic in simulations of physical robots. We agree that real robots and environments must be used during the design of simulators which attempt to model them. It is plausible, however, that once a simulator has been demonstrated to ....

....between them to trade off precision against speed in the simulation. The user should be able to control the granularity of the simulation in both space and time. Continuous motion simulation should be used wherever possible, as unintended effects from discrete simulation have been documented [5]. Pseudorandom sequences used to generate data for simulated sensors should be repeatable exactly. If the simulator has control over the robot control architecture, and if that architecture is deterministic, this will allow a repeatability not possible in the physical world that can reveal elusive ....

Rodney A. Brooks. Artificial life and real robots. In Proceedings of European Conference on Artificial Life, December 1991.

....to allow the autonomous acquisition of task achieving competences: the synthetic step takes place in the robot controller. This difference is, I believe, an important one. Brooks reports that it is not yet possible to autonomously acquire complete behaviours, using the subsumption architecture ( Brooks 91a, sect.1] Progress in learning new behaviours has proven difficult. Today, we are constrained to programming each new behaviour by hand. Alder and Cairngorm, however, are able to acquire new behaviours (even if the designer has not thought of them ) The dead end escape behaviour described ....

....to be recharged 4 certain competences such as obstacle avoidance have to be acquired very quickly in order to ensure safe operation of the robot. The conclusion is: for mobile robotics it is crucial that the learning algorithm is fast (on the slow speed of reinforcement learning see also [Brooks 91a] The fact that reinforcement learning can be extremely slow is shown by other researchers, too. Prescott Mayhew] simulate the AIVRU mobile robot and use a reinforcement learning algorithm similar to the one described by [Watkins 89] The sensor input space of the simulated agent is a ....

[Article contains additional citation context not shown here]

Rodney Brooks, Artificial Life and Real Robots, Proceedings of 1st European Conference on Artificial Life 1991, MIT Press Cambridge Mass. and London, England, 1991.

....sensory modality that has been extensively studied in both natural and artificial systems, and we believe visually guided agents should be studied from as early a stage as possible. ffl While we could impose on our robot some visual sensors with fixed properties, we advocate (in common with Brooks [9]) the concurrent evolution of visual sensor morphology and the control networks: separating morphology from control is a measure which is difficult to justify from an evolutionary perspective, and potentially misleading. ffl For reasons of parsimony, studies of visually guided agents should ....

R. A. Brooks. Artificial life and real robots. In F. J. Varela and P. Bourgine, editors, Toward a Practice of Autonomous Systems: Proceedings of the First European Conference on Artificial Life (ECAL91), pages 3--10, Cambridge MA, 1992. MIT Press Bradford Books.

....1992. 42] D. T. Cliff, Computational neuroethology: A provisional manifesto, in From Animals to Animats, First International Conference on Simulation of Adaptive Behavior (SAB 90) J. A. Meyer and e. S. Wilson, eds. pp. 29 39, MIT Press, 1991. 1] 2] 3] 4] 5] 6] 7] 8] 9] 10] [11] [12] 13] 14] 15] 16] 17] 18] 19] 20] 21] 22] 23] 24] 25] 26] 27] 28] 29] 30] 31] 32] 33] 34] 35] 36] 37] 38] 39] 40] 41] 42] 43] 44] 45] 46] 47] 48] 49] 50] 51] 52] 53] 54] 55] 56] 57] 58] 59] 60] 61] 62] 63] 64] 65] 66] 67] 68] 69] 70] ....

R. A. Brooks, "Artificial life and real robots," in Toward a Practice of Autonomous Systems: Proceedings of the First European Conference on Artificial Life (F. J. Varela and e. Paul Bourgine, eds.), pp. 3--10, MIT Press, 1992.

....from simulation to physical implementation of robotic agent experiments is a non trivial exercise. One of the recurrent themes in the literature is that the gap between simulation and physical realisation is so wide that results from simulation do not carry over at all to physical realisation [1]. Others have argued that the gap is not so wide and that results obtained in simulated environments (including results from experiments in deriving behavioral networks through genetic algorithms) can be relatively easily ported to physical robots [4] In the past years, we have been working on ....

Brooks, R. (1992) Artificial life and real robots. In: F.J. Varela and P. Bourgine (eds.) Toward a Practice of Autonomous Systems. Proceedings of the First European Conference on Artificial Life. MIT Press/Bradford Books, Cambridge Ma. p. 3-10.

.... not adequate for complex environments (specially when the animat starts with a set of simple actions in comparison with the needed to solve the tasks proposed in there) The other strategy is the evolution by means of selection and variation of the best adapted animats, proposed for instance in [3], 12] or [11] This approach usually states that animats are born completely hard wired (adapted) not by the design made by an engineer, but by the natural process of evolution simulated by genetic algorithms [10] This approach is borrowed from nature and reflects the fact that animals are born ....

R.A. Brooks. Artificial life and real robots. In Proceedings of ECAL91. MIT Press/Bradford Books, 1992.

....1993) An Marco Dorigo ALECSYS AND THE AUTONOMOUSE 4 architecture built using ALECSYS is composed of a (hierarchically structured) set of cooperating modules, each one implementing either a simple behavior pattern or a coordination activity. In this respect our work is related to that of Brooks (1990). A major difference is that Brooks behaviors are designed and not learned. Also closely related is the work of Mahadevan and Connell (1992) which uses a subsumption architecture (Brooks, 1991a) whose components are learning systems instead of being finite state machines, like in Brook s ....

Brooks, R.A. (1991b). Artificial life and real robots. Proceedings of the First European Conference on Artificial Life, Paris, MIT Press/Bradford Books, 3--10.

No context found.

Brooks, R. (1992). Artificial life and real robots. In Varela, F. and Bourgine, P., editors, Toward a practice of autonomous systems: Proceedings of the First European Conference on Artificial Life, pages 3--10. MIT Press.

No context found.

R. Brooks. Artificial life and real robots. In F. Varela and P. Bourgine, editors, Proceedings of the First European Conference on Artificial Life, pages 3--10. MIT Press, 1992.

No context found.

R. Brooks. Artificial life and real robots. European Conf. on Artificial Life, 1992.

No context found.

R. A. Brooks, "Artificial life and real robots," in Toward a Practice of Autonomous Systems: Proceedings of the First European Conference on Artificial Life, (Cambridge, MA), pp. 3--10, MIT Press, 1992.

No context found.

Brooks, R.A. "Artificial life and real robots." Towards a practice of autonomous systems : Proceedings of the first European conference on artificial life. Eds F.J. Varela and P. Bourgine. MIT press. 1991.

No context found.

R.A. Brooks, Artificial life and real robots, in: F.J. Varela, P. Bourgine (Eds.), Toward a Practice of Autonomous Systems: Proceedings of the First European Conference on Artificial Life, MIT Press/Bradford Books, Cambridge, MA, 1992, pp. 3--10.

No context found.

Rodney A. Brooks. Artificial life and real robots. In Francisco J. Varela and Paul Bourgine, editors, Towards a Practice of Autonomous Systems: Proceedings of the First European Conference on Artificial Life, pages 3--10, Cambridge, MA, 1992. MIT Press.

No context found.

Brooks R. A. (1991a). Artificial Life and Real Robots , Towards a Practice of Autonomous Systems: European Conference on Artificial Life, Paris, France, MIT Press, December, pp. 3 - 10.

No context found.

) R. A. Brooks. Artificial life and real robots. In Proc. of the Fifth European Conference on Artificial Life, pages 35--48, 1990.

No context found.

Brooks, R. (1992), "Artificial Life and Real Robots", in Varela, F. J. and P. Bourgine (eds.), Toward a Practice of Autonomous Systems. Proceedings of the First European Conference on Artificial Life (The MIT Press, Cambridge, MA), 3-10.

No context found.

Brooks R. Artificial Life and Real Robots. In Proceedings of the First European Conference on Artificial Life. F. Varela & P. Bourgine Eds., Cambridge, MIT Press, USA 1992.

First 50 documents  Next 50

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