Distributed Artificial Intelligence (DAI) has existed as a subfield of AI for less than two decades. DAI is

The goal of getting computers to automatically solve problems is central to artificial intelligence, machine learning, and the broad area encompassed by what Turing called “machine intelligence ” [161, 162].

RoboCup simulated soccer presents many challenges to reinforcement learning methods, including a large state space, hidden and uncertain state, multiple agents, and long and variable delays in the eects of actions. We describe our application of episodic SMDP Sarsa() with linear tile-coding function approximation and variable to learning higherlevel decisions in a keepaway subtask of RoboCup soccer. In keepaway, one team, \the keepers," tries to keep control of the ball for as long as possible despite the eorts of \the takers." The keepers learn individually when to hold the ball and when to pass to a teammate, while the takers learn when to charge the ball-holder and when to cover possible passing lanes. Our agents learned policies that signi cantly out-performed a range of benchmark policies. We demonstrate the generality of our approach by applying it to a number of task variations including dierent eld sizes and dierent numbers of players on each team.

1 RoboCup simulated soccer presents many challenges to reinforcement learning methods, in-cluding a large state space, hidden and uncertain state, multiple independent agents learning simultaneously, and long and variable delays in the effects of actions. We describe our appli-cation of episodic SMDP Sarsa(λ) with linear tile-coding function approximation and variable λ to learning higher-level decisions in a keepaway subtask of RoboCup soccer. In keepaway, one team, “the keepers, ” tries to keep control of the ball for as long as possible despite the efforts of “the takers. ” The keepers learn individually when to hold the ball and when to pass to a teammate. Our agents learned policies that significantly outperform a range of benchmark policies. We demonstrate the generality of our approach by applying it to a number of task variations including different field sizes and different numbers of players on each team.

Cooperative multi-agent systems are ones in which several agents attempt, through their interaction, to jointly solve tasks or to maximize utility. Due to the interactions among the agents, multi-agent problem complexity can rise rapidly with the number of agents or their behavioral sophistication. The challenge this presents to the task of programming solutions to multi-agent systems problems has spawned increasing interest in machine learning techniques to automate the search and optimization process. We provide a broad survey of the cooperative multi-agent learning literature. Previous surveys of this area have largely focused on issues common to specific subareas (for example, reinforcement learning or robotics). In this survey we attempt to draw from multi-agent learning work in a spectrum of areas, including reinforcement learning, evolutionary computation, game theory, complex systems, agent modeling, and robotics. We find that this broad view leads to a division of the work into two categories, each with its own special issues: applying a single learner to discover joint solutions to multi-agent problems (team learning), or using multiple simultaneous learners, often one per agent (concurrent learning). Additionally, we discuss direct and indirect communication in connection with learning, plus open issues in task decomposition, scalability, and adaptive dynamics. We conclude with a presentation of multi-agent learning problem domains, and a list of multi-agent learning resources. 1

This paper presents layered learning, a hierarchical machine learning paradigm. Layered learning applies to tasks for which learning a direct mapping from inputs to outputs is intractable with existing learning algorithms. Given a hierarchical task decomposition into subtasks, layered learning seamlessly integrates separate learning at each subtask layer. The learning of each subtask directly facilitates the learning of the next higher subtask layer by determining at least one of three of its components: (i) the set of training examples; (ii) the input representation; and/or (iii) the output representation. We introduce layered learning in its domainindependent general form. We then present a full implementation in a complex domain, namely simulated robotic soccer. 1. Introduction Machine learning (ML) algorithms select a hypothesis from a hypothesis space based on a set of training examples such that the chosen hypothesis is predicted to characterize unseen examples...

We present an adaptation of the standard genetic program (GP) to hierarchically decomposable, multi-agent learning problems. To break down a problem that requires cooperation of multiple agents, we use the team objective function to derive a simpler, intermediate objective function for pairs of cooperating agents. We apply GP to optimize first for the intermediate, then for the team objective function, using the final population from the earlier GP as the initial seed population for the next. This layered learning approach facilitates the discovery of primitive behaviors that can be reused and adapted towards complex objectives based on a shared team goal.

Genetic programming is an automatic method for creating a computer program or other complex structure to solve a problem. This paper first reviews various instances where genetic programming has previously produced human-competitive results. It then presents new human-competi- Z. tive results involving the automatic synthesis of the design of both the parameter values i.e., tuning and the topology of controllers for two illustrative problems. Both genetically evolved controllers are better than controllers designed and published by experts in the field of control using the criteria established by the experts. One of these two controllers infringes on a previously issued patent. Other evolved controllers duplicate the functionality of other previously patented controllers. The results in this paper, in conjunction with previous results, reinforce the prediction that genetic programming is on the threshold of routinely producing human-competitive results and that genetic programming can potentially be used as an "invention machine" to produce patentable new inventions.

Temporal difference (TD) learning (Sutton and Barto, 1998) has become a popular reinforcement learning technique in recent years. TD methods, relying on function approximators to generalize learning to novel situations, have had some experimental successes and have been shown to exhibit some desirable properties in theory, but the most basic algorithms have often been found slow in practice. This empirical result has motivated the development of many methods that speed up reinforcement learning by modifying a task for the learner or helping the learner better generalize to novel situations. This article focuses on generalizing across tasks, thereby speeding up learning, via a novel form of transfer using handcoded task relationships. We compare learning on a complex task with three function approximators, a cerebellar model arithmetic computer (CMAC), an artificial neural network (ANN), and a radial basis function (RBF), and empirically demonstrate that directly transferring the action-value function can lead to a dramatic speedup in learning with all three. Using transfer via inter-task mapping (TVITM), agents are able to learn one task and then markedly reduce the time it takes to learn a more complex task. Our algorithms are fully implemented and tested in the RoboCup soccer Keepaway domain. This article contains and extends material published in two conference papers (Taylor and Stone, 2005; Taylor et al., 2005).

Abstract. RoboCup simulated soccer presents many challenges to machine learning (ML) methods, including a large state space, hidden and uncertain state, multiple agents, and long and variable delays in the effects of actions. While there have been many successful ML applications to portions of the robotic soccer task, it appears to be still beyond the capabilities of modern machine learning techniques to enable a team of 11 agents to successfully learn the full robotic soccer task from sensors to actuators. Because the successful applications to portions of the task have been embedded in different teams and have often addressed different subtasks, they have been difficult to compare. We put forth keepaway soccer as a domain suitable for directly comparing different machine learning approaches to robotic soccer. It is complex enough that it can’t be solved trivially, yet simple enough that complete machine learning approaches are feasible. In keepaway, one team, “the keepers, ” tries to keep control of the ball for as long as possible despite the efforts of “the takers. ” The keepers learn individually when to hold the ball and when to pass to a teammate, while the takers learn when to charge the ball-holder and when to cover possible passing lanes. We fully specify the domain and summarize some initial, successful learning results. 1