P. Stone and M. Veloso. A layered approach to learning client behaviors in the robocup soccer server. Applied Artificial Intelligence, 12:165--188, 1998.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....of learning applications that are typically studied for multi agent and multi robot systems vary considerably in their characteristics. Some of the learning application domains include air fleet control [34] predator prey [4] 17] 14] box pushing [22] foraging [24] and multi robot soccer [35], 23] 16] Of the previous application domains that have been studied in the context of multi robot learning, only the multi robot soccer domain addresses inherently cooperative tasks with more than two robots while also addressing the real world complexities of embodied robotics, such as ....

P. Stone and M. Veloso. A layered approach to learning client behaviors in the robocup soccer server. Applied Artificial Intelligence, 12:165-188, 1998.

....systems, such as that found in higher animals including humans, is being studied in domains such as multi robot soccer. A previous special journal issue in Artificial Intelligence on RoboCup discusses many of the advances in this area; see [6] for a general overview of the field, and [49] 7] [61], 64] 5] for some particular examples of this research. Another series of books appears yearly in the Lecture Notes in Artificial Intelligence series on the topic of multi robot soccer, beginning with [38] Two articles in this special issue address multi robot control issues in the ....

P. Stone and M. Veloso. A layered approach to learning client behaviors in the robocup soccer server. Applied Artificial Intelligence, 12:165--188, 1998.

....has generated advances in cooperative control. Significant study in predator prey systems has occurred, although primarily in simulation [7,21] Competition in multi robot systems, such as found in higher animals including humans, is beginning to be studied in domains such as multi robot soccer [46,29]. These areas of biological inspiration and their applicability to multi robot teams seem to be fairly well understood. More recently identified, less well understood, biological topics of relevance include the use of imitation in higher For a more detailed analysis of various types of ....

....has been accomplished to date in multi robot learning. The types of applications that are typically studied for this area of multi robot learning vary considerably in their characteristics. Some of the applications include predator prey [7,21] box pushing [28] foraging [31] multi robot soccer [46,29,41], and cooperative target observation [37] Particularly challenging domains for multi robot learning are those tasks that are inherently cooperative that is, tasks in which the utility of the action of one robot is dependent upon the current actions of the other team members. Inherently ....

P. Stone and M. Veloso. A layered approach to learning client behaviors in the robocup soccer server. Applied Artificial Intelligence, 12:165-188, 1998.

....is shown in Table 2.2. In the bottom layer the ball interception behavior has been learned using a neural network. The pass evaluation behavior in the second layer has been learned using the C4.5 decision tree algorithm (see [75] and uses the learned ball interception skill from the layer below [93, 97]. Subsequently, the pass selection behavior in the third layer has been learned using a new multi agent reinforcement learning method called TPOT RL with the pass evaluation skill Team Partitioned Opaque Transition Reinforcement Learning: this method can be used for maximizing long term ....

....references we also summarize the significant results of these teams in international competitions (top 10 finishes only) It is important to realize that the list is not exhaustive and only contains references and results up to and including the year 2000. Team References Roll of honour CMUnited [7, 90, 92, 93, 94, 95, 96, 97] 4th WC97, 1st WC98, 98, 99, 100, 102, 103, 104, 115] 1st 9th WC99, 3rd 4th WC00 Essex Wizards [36, 37, 39, 50, 51, 52, 53] 3rd WC99, 3rd EC00, 7th WC00 FC Portugal [56, 76, 77] 1st EC00, 1st WC00 Cyberoos [13, 69, 70, 71, 72, 73, 74] 3rd PR98, 4th EC00, 9th WC00 Karlsruhe Brainstormers [79, ....

P. Stone and M. Veloso. A Layered Approach to Learning Client Behaviors in the RoboCup Soccer Server. Applied Artificial Intelligence, 12:165--188, 1998.

....be assigned less frequently. At the strategic level it is interesting to adapt the agents actions to the behavior of other intelligent entities (either artificial or human) present in the environment [13] In particular, this is useful whenever the application domain provides for hostile agents [14]. In this case the aim is to foresee what the other agents are going to do, so that the MAs can carry out a more effective activity allocation. The strategist, on the basis of the opponents behaviors, chooses the most suitable tactics for hindering them. 3.2 How to Adapt Even if the elements ....

P. Stone and M. Veloso, "A layered approach to learning client behaviors in the robocup soccer server," Applied Articial Intelligence Journal, vol. 12, 1998.

.... [KAK 95, As99] These have been found to be rich experimental environments for many MAS research areas, including flexible teamwork learning [TAA 99] and methodologies, including hierarchical sensing and reinforcement learning by Q learning, temporal differences, team partitioned algorithms [SV98], artificial neural networks, and genetic programming [As99] At present, however, hand coded and hybrid learning techniques that employ a large amount of hand coded domain specific knowledge still outperform strategies that are learned automatically. In keep away soccer, three offensive agents ....

P. Stone and M. Veloso. A Layered Approach to Learning Client Behaviors in the RoboCup Soccer Server. Applied Artificial Intelligence (AAI) 12(3):165-188. Taylor and Francis, London, UK, 1998.

....and local trajectory planning algorithms are presented. Soccer (association football of robots) is team game in which a players have a cooperation [1,7] This is a real time game where situation changes dynamically. Soccer was chosen as one of problems for studying on multi agent systems [9,10]. We are designing the soccer agent using the cognitive approach. We present the decision making algorithm for cooperative action among soccer players. We developed the learning modules for partial implementation of decision making at high and low behavior levels of soccer agent. This soccer agent ....

P. Stone, M. Veloso. A layered Approach to Learning Client Behaviors in RoboCup Soccer Server Applied Artificial Intelligence, 12, 1998.

....Hence, much current work is ongoing in the eld of multiagent learning (e.g. 48] The types of applications that are typically studied vary considerably in their characteristics. Some of the applications include air eet control [40] predator prey [3] 18] 16] 44] and multi robot soccer [43], 23] In section 7, we mentioned previous works that have investigated various issues using the box pushing application domain, which include [11] 42] 20] 25] 37] In [47] Weiss describes a system with similar goals to the learning described in this article namely, how agents can ....

P. Stone and M. Veloso. A layered approach to learning client behaviors in the robocup soccer server. Applied Articial Intelligence, 12:165-188, 1998.

.... (see, for instance (Sen, Sekaran, Hale 1994) and with communication (Rus, Donald, Jennings 1995; Mataric 1997; Sasaki et al. 1995) Other problems where cooperation is essential are those faced by teams where each member plays a different role, as, for instance, players on a soccer team (Stone Veloso 1997). Tambe (1997) has proposed methods for agent teams working in synthetic domains where individual team members have their own reactive plans, but are also given team plans that explicitly express the team s joint activities. 2. Problems where cooperation increases performance either by decreasing ....

Stone, P., and Veloso, M. M. 1997. A layered approach to learning client behaviors in the robocup soccer server. Journal of Applied Artificial Intelligence.

....angle (with respect to the current looking direction of the player) and the distance of the object. Disturbances are added to the position data and to the robot commands. This simulates some of the uncertainty found when using real robots. The simulation is useful to test concepts for learning [70] strategies. In particular because it allows a large number of experiments in short time. But it is di#erent from real robots since it does not attempt a detailed simulation of the robot sensors and e#ectors. Small size league The playground for this league has the size of a ping pong table. The ....

P. Stone and M. Veloso. Layered approach to learning client behaviors in the robocup soccer server. Applied Artificial Intelligence, 12(2--3):165--187, March--May 1998.

....Even systems that learn or reason, learn or reason better when enabled by good design. Both experience and formal arguments have shown that structure and bias are necessary to these processes, in order to reduce the search space of their learning or reasoning algorithms to a tractable size [8, 29, 34 36]. Agent technology is itself a major design innovation which harnesses the innate ability of human designers to reason about societies of actors [19] The importance of design explains the focus of our community on agent architectures. Agent architectures are essentially design methodologies: they ....

Peter Stone and Manuela Veloso. A layered approach to learning client behaviors in the robocup soccer server. International Journal of Applied Artificial Intelligence, 12:165--188, 1998.

....in multi robot teams. The types of applications that are typically studied for this area of multirobot learning vary considerably in their characteristics. Some of the applications include air eet control [9] predator prey [10, 11, 12] box pushing [13] foraging [14] and multi robot soccer [15, 16]. Particularly challenging domains for multi robot learning are those tasks that are inherently cooperative. By this, we mean that the utility of the action of one robot is dependent upon the current actions of the other team members. Inherently cooperative tasks cannot be decomposed into ....

P. Stone and M. Veloso. A layered approach to learning client behaviors in the robocup soccer server. Applied Articial Intelligence, 12:165-188, 1998.

.... thorough approach to incorporate learning into an agent architecture is that of Stone and Veloso, who used neural networks for a low level individual behavior ball interception, which was then incorporated, via a decision tree, into the learning of a high level collaborative behavior passing (Stone and Veloso 1998). They further suggested that the output of the decision tree could be used as the input of a higher level learning module, though this higher level has not been implemented. A common weakness of these layered learning approaches is that there was no clear relationship between individual behavior ....

....explore when building the interaction between layers. The overall architecture is based on a bottom up approach where basic behaviors are first learned at the lowest level. Each successive layer learns to choose between the previous behaviors through machine learning to form more complex behaviors (Stone and Veloso 1998, Balch 1998, Luke et al. 1998) Once the desired level of complexity for an individual agent is attained, the same procedure is used to form cooperative behaviors between multiple agents. This cooperative level of behavior can have as many layers as needed to achieve the appropriate level of team ....

[Article contains additional citation context not shown here]

Stone, P. and Veloso, M. 1998. A Layered Approach to Learning Client Behaviors in the RoboCup Soccer Server. Applied Artificial Intelligence 12:165-188.

....re exes, the visual system, etc. It is dicult to imagine a monolithic entity that would be capable of the range and complexity of behaviors that mammals exhibit. Similarly, hierarchical approaches have been proposed to help create agents for complex control tasks (e.g. 2, 4] Layered learning [19, 20] is such a hierarchical paradigm that relies on learning the various subtasks necessary for Submitted for conference review, November, 2002. achieving the complete high level goal. Layered learning is a bottom up paradigm by which low level behaviors (those closer to the environmental inputs) are ....

P. Stone and M. Veloso. A layered approach to learning client behaviors in the RoboCup soccer server. Applied Arti cial Intelligence, 12:165-188, 1998.

....environments. Our general approach, called layered learn ing, is based on the premise that realistic domains are too complex for learning mappings directly from sensor inputs to actuator outputs. Instead, intermediate domain dependent skills should be learned in a bottom up hierarchical fash ion [9]. We implemented TPOT RL as the current highest layer of a layered learning system in the RoboCup soccer server [7] The soccer server used at RoboCup 97 [4] is a much more complex domain than has previously been used for studying multi agent policy learning. With 11 players on each team ....

....a likely success (ball reaches its destination or a teammate gets it) or likely failure (opponent intercepts the ball) Using the C4.5 DT algorithm [8] the classifications were learned with associated confidence factors. The learned behaviors proved effective both in controlled testing scenarios [9, 11] and against other previously unseen opponents in an international tournament setting [4] These two previously learned behaviors were both trained off line in limited, controlled training situations. They could be trained in such a manner due to the fact that they only involved a few players: ....

[Article contains additional citation context not shown here]

Peter Stone and Manucla Vcloso. A layered approach to learning client behaviors in the RoboCup soccer server. Applied Artificial Intelligence, 12:165 188, 1998.

....Each layer is learned by applying an ML algorithm that is appropriate for the specific subtask characteristics. In this paper, we apply layered learning to a complex multiagent learning task, namely simulated robotic soccer. We have previously presented the individual learned tasks in this domain [18, 20] as well as a preliminary version of the concept of layered learning [18] This paper contributes the concrete domain independent specification of layered learning as presented in Sections 2 and 3. Section 4 reviews our machine learning research in the simulated robotic soccer domain, couching it ....

....specific subtask characteristics. In this paper, we apply layered learning to a complex multiagent learning task, namely simulated robotic soccer. We have previously presented the individual learned tasks in this domain [18, 20] as well as a preliminary version of the concept of layered learning [18]. This paper contributes the concrete domain independent specification of layered learning as presented in Sections 2 and 3. Section 4 reviews our machine learning research in the simulated robotic soccer domain, couching it in the terms of our layered learning specification. This layered learning ....

[Article contains additional citation context not shown here]

Peter Stone and Manuela Veloso. A layered approach to learning client behaviors in the RoboCup soccer server. Applied Artificial Intelligence, 12:165--188, 1998.

....and opponents. The agent uses a complex heuristic decision mechanism, incorporating a machine learning module, to choose its action. The most significant changes from CMUnited 98 are that the agents use special purpose code for breakaways (see Section 2. 4) that the pass evaluation decision tree [10] has been retrained during practice games to capture the agents improved ball interception ability (see Section 2.2; that the agents can cross the ball (see below) and that the agents consider whether there is a teammate in a better position than they are to shoot the ball (see Section 2.3) In ....

Peter Stone and Manuela Veloso. A layered approach to learning client behaviors in the RoboCup soccer server. Applied Artificial Intelligence, 12:165--188, 1998.

....planning. Along with the real robot competition, RoboCup97 will also include a simulator based tournament using the Soccer Server system designed by Noda [ Noda, 1995 ] While we continue working on our real world system, we have been concurrently developing learning techniques in simulation [ Stone and Veloso, 1997; 1996 ] We eventually hope to transfer these learning techniques to the real system as we develop a complete Robotic Soccer architecture. This paper describes the overall architecture of our robotic soccer system. The combination of robust hardware, real time vision, and intelligent control ....

....for such a purpose since the detection of ball s location is noisy. The Kalman filter takes into account the existence of such noise and gives a best estimate. 5 The Robot Control Code The robot control code itself consists of several different behavior levels as summarized in Figure 2 [ Stone and Veloso, 1997 ] First, there is reactive control code which enables the robot to move to a goal location either a fixed position on the field or a moving target such as the ball. This reactive control is embedded within an obstacle avoidance routine which allows the robots to avoid both teammates and ....

Peter Stone and Manuela Veloso. A layered approach to learning client behaviors in the robocup soccer server. To appear in Applied Artificial Intelligence (AAI) Journal, 1997.

....Each layer is learned by applying an ML algorithm that is appropriate for the speci c subtask characteristics. In this paper, we apply layered learning to a complex multiagent learning task, namely simulated robotic soccer. We have previously presented the individual learned tasks in this domain (Stone Veloso, 1998; Stone Veloso, 1999b) as well as a preliminary version of the concept of layered learning (Stone Veloso, 1998) This paper contributes the concrete domainindependent speci cation of layered learning as presented in Sections 2 and 3. Section 4 reviews our machine learning research in the ....

....this paper, we apply layered learning to a complex multiagent learning task, namely simulated robotic soccer. We have previously presented the individual learned tasks in this domain (Stone Veloso, 1998; Stone Veloso, 1999b) as well as a preliminary version of the concept of layered learning (Stone Veloso, 1998). This paper contributes the concrete domainindependent speci cation of layered learning as presented in Sections 2 and 3. Section 4 reviews our machine learning research in the simulated robotic soccer domain, couching it in the terms of our new layered learning speci cation. This layered ....

[Article contains additional citation context not shown here]

Stone, P., & Veloso, M. (1998). A layered approach to learning client behaviors in the RoboCup soccer server. Applied Articial Intelligence, 12, 165-188.

No context found.

P. Stone and M. Veloso. A layered approach to learning client behaviors in the robocup soccer server. Applied Artificial Intelligence, 12:165--188, 1998.

No context found.

P. Stone, M. Veloso, "Layered Approach to Learning Client Behaviors in the RoboCup Soccer Server", Applied Artificial Intelligence, 12:165-188, 1998.

No context found.

P. Stone, M. Veloso, "Layered Approach to Learning Client Behaviors in the RoboCup Soccer Server", Applied Artificial Intelligence, 12:165-188, 1998.

No context found.

P. Stone and M. Veloso. Layered approach to learning client behaviors in the robocup soccer server. Applied Artificial Intelligence, 12(2-3):165-- 188, 1998. 203

No context found.

P. Stone and M. Veloso. A layered approach to learning client behaviors in the robocup soccer server. Applied Arti cial Intelligence, 12:165-188, 1998.

No context found.

P. Stone and M. Veloso, \A layered approach to learning client behaviors in the robocup soccer server," Applied Articial Intelligence 12, pp. 165-188, 1998.

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