Multiagent learning is a key problem in AI. In the presence of multiple Nashequilibria, even agents with non-conflicting interests may not be able to learn an optimal coordination policy. The problem is exaccerbated ifthe agents do not know the game and independently receive noisy payoffs. So, multiagent reinforfcement learning involves two interrelated problems:identifying the game and learning to play.

Imitation can be viewed as a means of enhancing learning in multiagent environments. It augments an agent's ability to learn useful behaviors by making intelligent use of the knowledge implicit in behaviors demonstrated by cooperative teachers or other more experienced agents. We propose and study a formal model of implicit imitation that can accelerate reinforcement learning dramatically in certain cases. Roughly, by observing a mentor, a reinforcement-learning agent can extract information about its own capabilities in, and the relative value of, unvisited parts of the state space. We study two specific instantiations of this model, one in which the learning agent and the mentor have identical abilities, and one designed to deal with agents and mentors with different action sets. We illustrate the benefits of implicit imitation by integrating it with prioritized sweeping, and demonstrating improved performance and convergence through observation of single and multiple mentors. Though we make some stringent assumptions regarding observability and possible interactions, we briefly comment on extensions of the model that relax these restricitions.

Shaping has proven to be a powerful but precarious means of improving reinforcement learning performance. Ng, Harada, and Russell (1999) proposed the potential-based shaping algorithm for adding shaping rewards in a way that guarantees the learner will learn optimal behavior. In this note, we prove certain similarities between this shaping algorithm and the initialization step required for several reinforcement learning algorithms. More specifically, we prove that a reinforcement learner with initial Q-values based on the shaping algorithm's potential function make the same updates throughout learning as a learner receiving potential-based shaping rewards. We further prove that under a broad category of policies, the behavior of these two learners are indistinguishable. The comparison provides intuition on the theoretical properties of the shaping algorithm as well as a suggestion for a simpler method for capturing the algorithm's benefit. In addition, the equivalence raises previously unaddressed issues concerning the efficiency of learning with potential-based shaping.

In common-interest stochastic games all players receive an identical payoff. Players participating in such games must learn to coordinate with each other in order to receive the highest-possible value. A number of reinforcement learning algorithms have been proposed for this problem, and some have been shown to converge to good solutions in the limit. In this paper we show that using very simple model-based algorithms, much better (i.e., polynomial) convergence rates can be attained. Moreover, our model-based algorithms are guaranteed to converge to the optimal value, unlike many of the existing algorithms.

We propose a new approach to reinforcement learning which combines least squares function approximation with policy iteration. Our method is model-free and completely off policy. We are motivated by the least squares temporal difference learning algorithm (LSTD), which is known for its efficient use of sample experiences compared to pure temporal difference algorithms. LSTD is ideal for prediction problems, however it heretofore has not had a straightforward application to control problems. Moreover, approximations learned by LSTD are strongly influenced by the visitation distribution over states.

Tangential hand velocity profiles of rapid human arm movements often appear as sequences of several bell-shaped acceleration-deceleration phases called submovements or movement units. This suggests how the nervous system might efficiently control a motor plant in the presence of noise and feedback delay. Another critical observation is that stochasticity in a motor control problem makes the optimal control policy essentially different from the optimal control policy for the deterministic case. We use a simplified dynamic model of an arm and address rapid aimed arm movements. We use reinforcement learning as a tool to approximate the optimal policy in the presence of noise and feedback delay. Using a simplified model we show that multiple submovements emerge as an optimal policy in the presence of noise and feedback delay. The optimal policy in this situation is to drive the arm's end point close to the target by one fast submovement and then apply a few slow submovements to accurately drive the arm's end point into the target region. In our simulations, the controller sometimes generates corrective submovements before the initial fast submovement is completed, much like the predictive corrections observed in a number of psychophysical experiments.

We report on the use of reinforcement learning with Cobot, a software agent residing in the well-known online community LambdaMOO. Our initial work on Cobot (Isbell et al.2000) provided him with the ability to collect social statistics and report them to users. Here we describe an application of RL allowing Cobot to take proactive actions in this complex social environment, and adapt behavior from multiple sources of human reward. After 5 months of training, and 3171 reward and punishment events from 254 different LambdaMOO users, Cobot learned nontrivial preferences for a number of users, modifing his behavior based on his current state. Here we describe LambdaMOO and the state and action spaces of Cobot, and report the statistical results of the learning experiment.

We consider the use of two additive control variate methods to reduce the variance of performance gradient estimates in reinforcement learning problems. The first approach we consider is the baseline method, in which a function of the current state is added to the discounted value estimate. We relate the performance of these methods, which use sample paths, to the variance of estimates based on iid data. We derive the baseline function that minimizes this variance, and we show that the variance for any baseline is the sum of the optimal variance and a weighted squared distance to the optimal baseline. We show that the widely used average discounted value baseline (where the reward is replaced by the difference between the reward and its expectation) is suboptimal.

Competition between actions is based on the motivating characteristics of their consequent states in this sense.

However, online RL algorithms such as SARSA(A) have been shown experimentally to have difficulty converging when applied with function approximators. Theoretical analysis has not been able to prove convergence, even in the case-of linear function approximators.