P. Stone and M. Veloso, Multiagent Systems: A Survey from a Machine Learning Perspective, Journal of Autonomous Robots, Vol. 8, No. 3, pp 345-383, 2000.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....of Artificial Life (ALife) is to understand systems that possess life (i.e. survive, adapt and reproduce) by modeling them and the environment they are situated in. Multiagent systems differ from single agent systems in that several agents exist which model each other s goals and actions [4]. From a single agent s perspective, a MAS differs from a single agent system in that the environment s dynamics can be determined by other agents. One of the areas of intense interest by researchers is that of multi agent simulation. By using a MAS we have the possibility of directly ....

....areas of intense interest by researchers is that of multi agent simulation. By using a MAS we have the possibility of directly representing individuals and their interaction in a simulation. What characterises a MAS from other approaches is the dynamics of interaction between agents in the system [1, 3, 4] It is the intrinsic nature of multi agent systems that is driving researchers to apply them as simulation techniques, when building models of systems viewed as non linear and complex. A complex system is one in which an algorithm could describe the behaviour, but where mathematical models do ....

P. Stone and M. Veloso, "Multi-agent Systems: A Survey from a Machine Learning Perspective", Carnegie Mellon University, School of Computer Science, Pittsburgh, PA CMU-CS-97-193, 1997.

....in Multi Robot Systems The multi agent pursuit evasion game scenario considered in this paper fits within the general framework of multi robot systems. There exists a large body of literature in multi robot systems addressing problems such as machine learning techniques for multi agent systems [14], hybrid algorithms for multi agent control [15] multi robot localization [1] 16] distributed sensor fusion [17] and formation control [18] As for application of multi robot systems in robot soccer, we refer the reader to [19] and [20] for centralized coordination and control of multiple ....

....longer to capture a faster random evader than a slower random evader. This is because a faster random evader visits more cells in the map, increasing the chances of being detected. In Figure 5, for example, the higher speed of E1 allows it to move away from the visibility region of P2 for t [0, 14], but E1 soon moves into the visibility region of P3 and is quickly captured. UAV Pursuer vs. UGV Pursuer: Simulation results in [30] and Experiment 4 show that the local max policy has a similar performance with either a UAV or UGV pursuer, while the global max policy performs better with a ....

P. Stone and M. Veloso, "Multiagent systems: A survey from a machine learning perspective," Autonomous Robots, vol. 8, no. 3, pp. 345--383, 2000.

....agent to environment interactions, performance is often stochastic, and evaluative, rather than gradient based, search methods are appropriate. This type of control optimization has been extensively studied for the case of a single agent [14] 15] 16] as well as for multiple agents [17] 18] [19]. II. The Flocking Task A. Task Definition The flocking task examined in this paper is similar in form to the cooperative movement task studied in [20] The agents begin each trial at random positions and orientations within an area A located in the corner of a square arena of length L. During ....

P. Stone and M. Veloso, "Multiagent systems: A survey from a machine learning perspective," Autonomous Robots, vol. 8, no. 3, 2000.

....at nding dominant strategies. They represent a exible and general testbed for the analysis of dynamic properties of multiagent systems. 3. 4 The Predator Prey ( Pursuit ) domain The Predator Prey or Pursuit domain (pursuit domain in the following was introduced in [BJD86] and it is used in [SV00] to illustrate the various features of Multi Agent Systems (MAS) Even if it is a toy domain, it can be used for testing the basic features of a Society of Computees (SoC) In this domain there is a number of predators and a prey that move on a 2 dimensional world (see gure 1a) The predators ....

....landscapes and frustration, dissemination of culture, etc. More applications can be found following the links from the UMBC Agents web page, maintained by Tim Finin: http: agents.umbc.edu; they mostly refer to the e commerce context. 4 Communication in Multi agent Systems: state of the art In [SV00] a survey on the state of the art in MAS is presented. The authors divide the systems in four classes: homogeneous non communicating agents, heterogeneous non communicating agents, homogeneous communicating agents and heterogeneous communicating agents. We will discuss in the following the last ....

[Article contains additional citation context not shown here]

Peter Stone and Manuela M. Veloso. Multiagent systems: A survey from a machine learning perspective. Autonomous Robots, 8(3):345-383, 2000.

....consisting of several components which work together towards a common goal. MAS on the other hand aims to provide principles for the construction of complex systems containing multiple independent agents and focuses on behavior management issues (e.g. coordination of behaviors) in such systems [101]. Since robotic soccer is an example of a multi agent domain, we are mainly interested in the latter of these two subdisciplines throughout this thesis. An agent can be seen as anything that is situated in an environment and that perceives this environment through sensors and acts upon it through ....

....advantages. Some domains even require the use of MAS as a discipline. For example, in cases where there are di#erent entities (think of people, organizations, etc. with di#erent (possibly conflicting) goals and proprietary information, a multi agent system is necessary to model their interactions [101]. But even in domains which do not necessarily require MAS, their use can bring several advantages: The presence of multiple agents can provide a method for parallel computation, thereby speeding up the operation of the system. This is especially the case for domains in which the overall task ....

[Article contains additional citation context not shown here]

P. Stone and M. Veloso. Multi-Agent Systems: A Survey from a Machine Learning Perspective. Autonomous Robotics, 8(3), July 2000.

....to be jointly applied. Indeed, since different predictors might present different performances on the same prediction problem, it is necessary to choose the set of tools having better performances in the prediction problem of interest for the user. In this paper we propose a Multi Agent system [2, 3, 5, 7] for supporting users in the process of predicting the three dimensional structure of proteins and, in a way completely transparent to the user, to automatically perform the following tasks: i) selecting the team of predictors to be jointly applied on the prediction problem of interest for the ....

P. Stone and M. Veloso. Multiagent systems: A survey from a machine learning perspective. J. of Autonumous Robotics, 8(3):345-383, 2000.

....be adopted by the client. After each round, a brief report is given out to the client listing the final profit made t , and the profit made by following agent s advice t . Clients are expected to compare t and In MAS, reinforcement learning techniques have been studied extensively (e.g. [18, 20]) User modeling Portfolio selections User evaluations reinforcement learning by broker agents Stock info. t . Here, the measure of trust is based on the deviation of client s decision towards agent s advice. For instance, if agent s opinion is to buy 5000 worth of stocks A and ....

P. Stone and M. Veloso, Multiagent systems: a survey from a machine learning perspective, Autonomous Robots, 8 (3), 2000.

....others goals and actions. From an individual agent s perspective, multiagent systems vary from single agent systems most signi cantly because the environment s dynamics can be affected by other agents. In addition to the uncertainty of a system, other agents intentionally a ect the environment [21]. In the case of multiple agents, each learning simultaneously, one particular agent is learning the value of actions in a non stationary environment. Thus, the convergence of the aforementioned Q learning algorithm is not necessarily guaranteed in a multiagent setting. 2 Game Theory In the ....

P. Stone and M. Veloso. Multiagent systems: A survey from a machine learning perspective, 1997.

....the paths of agents 1, 3, and 4, respectively. The simulation and visualization were run for about 30 time steps to collect the screen capture. Several MAS variants of robotic soccer exist; keep away soccer belongs to the category of multi agent learning with homogeneous, noncommunicating agents [SV00b] those that share identical code but have no direct channels of communication other than by observing the behavior of teammates. This type of problem requires more robust, autonomous solutions and is therefore an interesting framework for teamwork learning. Soccer, whether analyzed as a human ....

P. Stone and M. Veloso. Multiagent Systems: A Survey from a Machine Learning Perspective. Autonomous Robots, 8(3): 345-383. Kluwer Academic Publishers, Norwell, MA, 2000.

....field of multiagent systems is already very well known and spans a wide area of possible applications, there are several introductory publications that can be recommended for getting more information about this subject. Schmutter, 2000b] is a short introduction to the main ideas and approaches. [Stone and Veloso, 2000] is more detailed concentrating on machine learning methods. Weiss, 1999] finally is a book extensively covering most current MAS approaches and problems. The research field of multiagent systems has come into being as a subfield of distributed artificial intelligence which itself is a subfield ....

Peter Stone and Manuela Veloso. Multiagent systems: A survey from a machine learning perspective. Autonomous Robots, 8(3):345--383, 2000.

....action decomposition class of problem. As our framework makes no assumption on the transition and reward functions of the system, it should also be well suited to problems of policy decomposition. In fact, we have also tested our approach on a modified version of the prey predator problem (see [8]) where the predators must also avoid obstacles in their environment. Because of the limited perceptions of the agents, the policy of one agent alters the transitions probability of the other agents. Actually, even the basic behaviors are more difficult to learn and a special kind of incremental ....

P. Stone and M. Veloso. Multiagent systems: a survey from a machine learning perspective. Autonomous Robotics, 8(3), 2000.

....to a 2 . The propert below t the robot indeed goes nd left infinitel often. Proposition 2. #right ##left) 3. 3 Multi agent Systems: The Pursuit Game Example tor Pre (or Pursuit) wide riet of a proa hes nd it s ma di#erent insta tia ions thaca be used e di#erent multi afi t scenaSyI [25]. As the Zigza:fiyI exa:fiy nces of the tor Pre ve been modeled utonomous robots [16] Here we modela simple nce of this me. tors a d pre moveaL nd discrete, grid like toroida l world withsqua respa ces; n move o# one end of the boa rda comeba ck on the other end. Preda torsa nd pre move simulta ....

P. Stone and M. Veloso. Multiagent Systems: A Survey from a Machine Learning Perspective. Autonomous Robots, 8:345--383, 2000.

....Proposition 8. pGoZigzagq ( right left) 7. 4 Multi Agent Systems: The Pursuit Game Example The Predator Prey (or Pursuit) game [5] has been studied using a wide variety of approaches [17] and it has many di erent instantiations that can be used to illustrate di erent multi agent scenarios [38]. As the Zigzagging example, instances of the Predator Prey game have been modeled using autonomous robots [26] Here we model a simple instance of this game. The predators and prey move around in a discrete, grid like toroidal world with square spaces; they can move o one end of the board and ....

P. Stone and M. Veloso. Multiagent systems: A survey from a machine learning perspective. Autonomous Robots, 8:345-383, 2000.

....ant foraging behavior [11] 10] provides an overview of such biologically based artificial navigation systems. The construction task can serve as another useful test bed for multi agent systems (existing test beds include the robotic soccer environment [2] and the Predator Prey pursuit domain [9]) The ESMs used in our system are similar in principle to evidence grids developed in [8] which were tested on physical robots. 7] shows how spatial representation can be integrated into a behavior based architecture. Since the internal representation only stores the information necessary for ....

Peter Stone and Manuela Veloso. Multiagent systems: A survey from a machine learning perspective. Autonomous Robots, 8(3), June 2000.

....the action decomposition class of problem. As our framework makes no assumption on the transition and reward functions of the system, it should also be well suited to problems of policy decomposition. In fact, we have also tested our approach on a modi ed version of the prey predator problem (see [15]) where the predators must also avoid obstacles in their environment. Because of the limited perceptions of the agents, the policy of one agent alters the transitions probability of the other agents. In fact, even the basic behaviors are more dicult to learn and a special kind of incremental ....

P. Stone and M. Veloso. Multiagent systems: a survey from a machine learning perspective. Autonomous Robotics, 8(3), 2000.

....of such a system also avoids some of the main difficulties with metamorphic systems in that planning complexity is not directly proportional with the number of modules. Placing the modules on separate robots gives a natural hierarchy that simplifies control and repair. Multi robot research [6][9][10] can then be applied to the control of the team as a whole. Perhaps the most appealing feature of this type of system is that redundant components are actually in use prior to a failure. Therefore, no components are sitting idle waiting to take over when another component fails and no ....

P. Stone and M. Veloso. Multiagent systems: A survey from a machine learning perspective. In Autonomous Robots, Volume 8, number 3, July 2000.

....auction schemes for optimizing negotiation. Many groups have investigated learning methods applied to multiagent systems. For example, Prasad et al. 54d# Hh#h. v b43, 45] Haynes et al. 31] Balch [5] Tan [79] Schmidhuber [64] Schaerf et al. 63] Zeng and Sycara [87, 86] Stone and Veloso [73], Schneider et al. 65] Wei [83] and Brauer and Wei [9] Others have pursued methods of coalition formation to optimize multiagent coordination. Examples are work done by Sandholm and Lesser [58] Sandholm et al. 60] Zlotkin and Rosenschein [89] and Shehory and Kraus [68] Tambe [77] proposes ....

Stone, P., and Veloso, M., "Multiagent Systems: A Survey from a Machine Learning Perspective", Carnegie Mellon University Computer Science Technical Report, CMU-CS-97-193, 1997.

....networks, industrial process control and electric power distribution. Specific instances of successful practical application of MAS in these and other areas are summarized in the excellent work by Parunak [13] Some motivations for MAS solution to problems and a taxonomy scheme are presented in [14]. 3. The Hybrid Intelligent Control Agent Systems of interest in this work are typically governed by continuous time dynamic equations within particular modes and by discrete transitions when a mode change occurs. For instance, an Unmanned Air Vehicle (UAV) modes might include take off, hover, ....

P. Stone and M. Veloso, "Multi-Agent systems: A Survey from A Machine Learning Perspective," Autonomous Robots, vol. 8, no. 3, pp. 345--383, July 2000.

.... of a system should take into account all possible reorganizations due to adaptations to the environment which becomes quite impossible in massive MAS [3] or real world applications (e.g. collective robotics in open environments [19] The use of multi agent machine learning (ML) techniques [24] is not always of a sucient help in that way, since most of them are concerned in setting parameters in xed agent behaviors. The complexity of these behaviors is a constitutive part of the model: the behaviors can be ne tuned through more or less sophisticated techniques, but they cannot change ....

P. Stone and M. Veloso, Multiagent systems: A survey from machine learning perspective, Autonomous Robots, 8 (2000).

....level of the game, or use them to solve particular subtasks (for instance, the pass to a teammate [Asada et al. 1999] Therefore, the tendency in adaptive multi agent simulations is to study much simpler application domains. The Prey Predator pursuit domain involving several predators [Stone and Veloso, 1997] is such a benchmark and illustrates this trend. But in these latter cases, the problem is often oversimpli ed: the agents move in a grid world, they have few possible actions. Hence, the problem lacks 1 In order to make clear that we sometimes use the Classi er Systems formalism without applying ....

Stone, P. and Veloso, M. (1997). Multiagent systems: A survey from a machine learning perspective. Technical Report CMU-CS-97-193, School of Computer Science, Carnegie Mellon University, Pittsburg, PA 15213.

No context found.

P. Stone and M. Veloso. Multiagent systems: A survey from a machine learning perspective. Autonomous Robots, 8(3):345-- 383, July 2000.

No context found.

P. Stone and M. Veloso. Multiagent systems: A survey from a machine learning perspective. Autonomous Robots, 8(3):345-- 383, July 2000.

....and control of those systems as a new research area, which gathers knowledge from many elds (e.g. Distributed Systems, Computer Science, Arti cial Intelligence, Robotics and Control) When compared with single agent systems, MARS o er a number of advantages. According to Stone and Veloso [60] and Arkin [6] those include: Robustness Being distributed and parallel systems by nature, MARS are less prone to failures. Even if an agent is damaged, the overall performance will not be compromised (provided that there is redundancy) Scalability MARS have the potential to be open ....

.... for the generalization of new knowledge and reorganizes system information (Arkin [6] In MARS, learning can be used at di erent levels, from the lower level (single robot control parameters) to the high level control parameters (population control parameters) Stone [59] and Stone and Veloso [60] addressed this question and presents a study on the application of multi layer machine learning within multiagent systems, namely, the robotic soccer domain. Other approaches to learning in MARS include Balch work [7] Up to this point, the reinforcement learning machine learning technique has ....

Peter Stone and Manuela Veloso. Multiagent Systems: A Survey from a Machine Learning Perspective. Technical Report CMU-CS-97-193, CMU, School of Computer Science, Carnegie Mellon University, May 1997.

....an agent s entire internal state. However it might indicate the role that the agent is currently filling within the team strategy and any other particularly useful information as determined during the locker room agreement. 3. 6 Related Work Most inter agent communication models (as surveyed in [9]) assume reliable point to point message passing with negligible communication costs. In particular, KQML assumes point to point message passing, possibly with the aid of facilitator agents [3] Nonetheless, KQML performatives could be used for the content portions of our proposed communication ....

Peter Stone and Manuela Veloso. Multiagent systems: A survey from a machine learning perspective. Technical Report CMU-CS-97-193, Computer Science Department, Carnegie Mellon University, Pittsburgh, PA, December 1997.

....adapting, learning in the presence of multiple learners can be viewed as a problem of a moving target, where the optimal policy may be changing while we learn. Multiple approaches to multiagent learning have been pursued with different degrees of success (as surveyed in [Wei and Sen, 1996] and [Stone and Veloso, 2000]) Previous learning algorithms either converge to a policy that is not optimal with respect to the other player s policies, or they may not converge at all. In this paper we contribute an algorithm to overcome these shortcomings. We examine the multiagent learning problem using the framework of ....

P. Stone and M. Veloso. Multiagent systems: A survey from a machine learning perspective. Autonomous Robots, 8(3), 2000.

No context found.

P. Stone and M. Veloso, Multiagent Systems: A Survey from a Machine Learning Perspective, Journal of Autonomous Robots, Vol. 8, No. 3, pp 345-383, 2000.

No context found.

P. Stone and M. Veloso, "Multiagent systems: A survey from a machine learning perspective," Autonomous Robots, vol. 8, no. 3, pp. 345--383, 2000.

No context found.

P. Stone and M. Veloso. Multi-agent systems: A survey from a machine learning perspective. Autonomous Robots, 8(3):345--383, 2000.

No context found.

Stone, P., Veloso, M.: Multiagent Systems: A Survey from a Machine Learning Perspective. Autonomous Robotics, 8(3):345-383 (2000)

No context found.

P. Stone and M. Veloso. Multiagent systems: A survey from a machine learning perspective. IEEE Transactions on Knowledge and Data Engineering, 1996.

No context found.

P. Stone and M. Veloso, "Multiagent systems: a survey from a machine learning perspective," Autonomous Robots, vol. 8, pp. 345--383, 2000.

No context found.

P. Stone and M. Veloso. Multiagent systems: A survey from a machine learning perspective. Autonomous Robots, 8(3), 2000.

No context found.

P. Stone and M. Veloso. Multiagent systems: A survey from a machine learning perspective. Technical Report CMU-CS97 -193, School of Computer Science, Carnegie Mellon University, 1997.

No context found.

Stone, P. and Veloso, M. (1997). Multiagent systems: A survey from a machine learning perspective. Technical report, Carnegie Mellon University.

No context found.

Stone, P., Veloso, M.: Multiagent Systems: A Survey from a Machine Learning Perspective. Autonomous Robotics, 8(3):345-383 (2000)

No context found.

P. Stone and M. Veloso. Multiagent systems: a survey from a machine learning perspective. Autonomous Robotics, 8(3), 2000.

No context found.

Stone, P. and Veloso, M. (2000). Multi-Agent Systems: A Survey from a Machine Learning Perspective. Autonomous Robotics, 8(3).

No context found.

P. Stone and M. Veloso. Multiagent systems: A survey from a machine learning perspective. Autonomous Robots, 8:345--383, 2000.

No context found.

P. Stone and M. Veloso. Multiagent systems: A survey from a machine learning perspective. Report Series no CMU-CS-97-193, Carnegie Mellon University, Computer Science Department, Carnegie Mellon University, Pittsburgh, PA, 1997.

No context found.

P. Stone and M. Veloso. Multiagent systems: A survey from a machine learning perspective. Autonomous Robotics, 8(3), July 2000.

No context found.

Peter Stone and Manuela Veloso. Multi agent systems: Survey from a machine learning perspective. Technical Report Technical Report CMU-CS-97-193, 29 pages., CMU-RI, 1997.

No context found.

P. Stone and M. Veloso. Multiagent systems: A survey from a machine learning perspective. Technical Report 193, Carnegie Mellon University, December 1997. 33

No context found.

Peter Stone and Manuela Veloso. "Multiagent Systems: A Survey from a Machine Learning Perspective". Autonomous Robotics, Vol 8, Num 3, July, 2000.

No context found.

P. Stone, M. Veloso (1997) Multiagent System: A Survey from a Machine Learning Perspective, CMU CS Technical Report

No context found.

P. Stone and M. Veloso. Multiagent systems: A survey from a machine learning perspective. Autonomous Robots, 3(8):345--383, June 2000.

No context found.

P. Stone, M. Veloso (1997) Multiagent System: A Survey from a Machine Learning Perspective, CMU CS Technical Report.

No context found.

P. Stone and M. Veloso, "Multiagent Systems: A Survey from a Machine Learning Perspective," Autonomous Robots, 8(3), July 2000. 10

No context found.

P. Stone and M. Veloso. Multiagent systems: A survey from a machine learning perspective. Autonomous Robots, 8(3), 2000.

No context found.

P. Stone and M. Veloso. Multiagent systems: A survey from a machine learning perspective. J. of Autonumous Robotics, 8(3):345--383, 2000.

No context found.

Peter Stone, Manuela Veloso. Multi-agent Systems: A Survey from a Machine Learning Perspective. In Autonomous Robotics, volume 8, number 3, july 2000. 4

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