This paper proposes a new reinforcement learning (RL) paradigm that explicitly takes into account input disturbance as well as modeling errors. The use of environmental models in RL is quite popular for both off-line learning by simulations and for online action planning. However, the difference between the model and the real environment can lead to unpredictable, often unwanted results. Based on the theory of H oocontrol, we consider a differential game in which a'disturbing' agent (disturber) tries to make the worst possible disturbance while a'control' agent (actor) tries to make the best control input. The problem is formulated as finding a minmax solution of a value function that takes into account the norm of the output deviation and the norm of the disturbance.

Many algorithms for approximate reinforcement learning are not known to converge. In fact, there are counterexamples showing that the adjustable weights in some algorithms may oscillate within a region rather than converging to a point. This paper shows that, for two popular algorithms, such oscillation is the worst that can happen: the weights cannot diverge, but instead must converge to a bounded region. The algorithms are SARSA(O) and V(O); the latter algorithm was used in the well-known TD-Gammon program. 1 Introduction Although there are convergent online algorithms (such as TD()') [1]) for learning the parameters of a linear approximation to the value function of a Markov process, no way is known to extend these convergence proofs to the task of online approximation of either the state-value (V*) or the action-value (Q*) function of a general Markov decision process. In fact, there are known counterexamples to many proposed algorithms.

In multi-service communications networks, such as Asynchronous Transfer Mode (ATM) networks, resource control is of crucial importance for the network operator as well as for the users. The objective is to maintain the service quality while maximizing the operator's revenue. At the call level, service quality (Grade of Service) is measured in terms of call blocking probabilities, and the key resource to be controlled is bandwidth. Network routing and call admission control (CAC) are two such resource control problems. Markov decision processes offer a framework for optimal CAC and routing [1]. By modelling the dynamics of the network with traffic and computing control policies using dynamic programming [2], resource control is optimized. A standard assumption in such models is that calls arrive according to Poisson processes. This makes the models of the dynamics relatively simple. Although the Poisson assumption is valid for most user-initiated requests in communications networks, a number of studies [3, 4, 5] indicate that many types of arrival similar.

We present an expressive agent design language for reinforcement learning that allows the user to constrain the policies considered by the learning process.The language includes standard features such as parameterized subroutines, temporary interrupts, aborts, and memory variables, but also allows for unspecified choices in the agent program. For learning that which isn't specified, we present provably convergent learning algorithms. We demonstrate by example that agent programs written in the language are concise as well as modular. This facilitates state abstraction and the transferability of learned skills. 1 Introduction The field of reinforcement learning has recently adopted the idea that the application of prior knowledge may allow much faster learning and may indeed be essential if realworld environments are to be addressed. For learning behaviors, the most obvious form of prior knowledge provides a partial description of desired behaviors. Several languages for partial descriptions have been proposed, including Hierarchical Abstract Machines (HAMs) [8], semi-Markov options [12], and the MAXQ framework [4]. This paper describes extensions to the HAM language that substantially increase its expressive power, using constructs borrowed from programming languages. Obviously, increasing expressiveness makes it easier for the user to supply whatever prior knowledge is available, and to do so more concisely.

Substantial data support a temporal difference (TO) model of dopamine (OA) neuron activity in which the cells provide a global error signal for reinforcement learning. However, in certain circumstances, OA activity seems anomalous under the TO model, responding to non-rewarding stimuli. We address these anomalies by suggesting that OA cells multiplex information about reward bonuses, including Sutton's exploration bonuses and Ng et al's non-distorting shaping bonuses. We interpret this additional role for OA in terms of the unconditional attentional and psychomotor effects of dopamine, having the computational role of guiding exploration. 1 Introduction Much evidence suggests that dopamine cells in the primate midbrain play an important role in reward and action learning. Electrophysiological studies support a theory that OA cells signal a global prediction error for summed future reward in appetitive conditioning tasks (Montague et al, 1996; Schultz et al, 1997), in the form of a temporal difference prediction error term.

A key challenge for reinforcement learning is scaling up to large partially observable domains. In this paper, we show how a hierarchy ofbehaviors can be used to create and select among variable length short-term memories appropriate for a task. At higher levels inthe hierarchy, the agent abstracts over lower-level details and looks back over a variable number of high-level decisions in time. We formalize this idea in a framework called Hierarchical Suffix Memory (HSM). HSM uses a memory-based SMDP learning method to rapidly propagate delayed reward across long decision sequences.

Many algorithms for approximate reinforcement learning are not known to converge. In fact, there are counterexamples showing that the adjustable weights in some algorithms may oscillate within a region rather than converging to a point. This paper shows that, for two popular algorithms, such oscillation is the worst that can happen: the weights cannot diverge, but instead must converge to a bounded region. The algorithms are SARSA(O) and V(O); the latter algorithm was used in the well-known TD-Gammon program. 1 Introduction Although there are convergent online algorithms (such as TD()') [1]) for learning the parameters of a linear approximation to the value function of a Markov process, no way is known to extend these convergence proofs to the task of online approximation ofeither the state-value (V*) or the action-value (Q*) function of a general Markov decision process. In fact, there are known counterexamples to many proposed algorithms.For example, fitted value iteration can diverge even for Markov processes [2]; Q-Iearning with linear function approximators can diverge, even when the states are updated according to a fixed update policy [3]; and SARSA(O) can oscillate between multiple policies with different value functions [4].

We present an expressive agent design language for reinforcement learning thatallows the user to constrain the policies considered by the learning process.Thelanguage includes standard features such as parameterized subroutines,temporary interrupts, aborts, and memory variables, but also allows for unspecified choices in the agent program. For learning that which isn't specified, we present provably convergent learning algorithms. Wedemonstrate by example that agent programs written in the language are concise as well as modular. This facilitates state abstraction and the transferability of learned skills. 1 Introduction The field of reinforcement learning has recently adopted the idea that the application of prior knowledge may allow much faster learning and may indeed be essential if realworld environmentsare to be addressed. For learning behaviors, the most obvious form of prior knowledge provides a partial description of desired behaviors. Several languages for partial descriptions have been proposed, including Hierarchical Abstract Machines (HAMs) [8], semi-Markov options [12], and the MAXQ framework [4]. This paper describes extensions to the HAM language that substantially increase its expressive power,using constructs borrowed from programming languages. Obviously, increasing expressivenessmakes it easier for the user to supply whatever prior knowledge is available, and to do so more concisely.

Visual input, providedby a video camera on a miniature robot, is preprocessed by a set of Gabor filters on 31 nodes of a log-polar retinotopic graph. Unsupervised Hebbianlearning is employed to incrementally build a population of localized overlapping place fields. Place cells serve as basis functions forreinforcement learning. Experimental results for goal-oriented navigation of a mobile robot are presented.

Substantial data support a temporal difference (TO) model of dopamine (OA) neuron activity in which the cells provide a global error signal for reinforcement learning. However, in certain circumstances, OAactivity seems anomalous under the TO model, responding to non-rewarding stimuli. We address these anomalies bysuggesting that OA cells multiplex information about reward bonuses,including Sutton's exploration bonuses and Ng et al's non-distorting shaping bonuses. We interpret this additional role for OA in terms of the unconditional attentional and psychomotor effectsof dopamine, having the computational role of guiding exploration. 1 Introduction Much evidence suggests that dopamine cells in the primate midbrain play an important rolein reward and action learning. Electrophysiological studies support a theory that OA cells signal a global prediction error for summed future reward in appetitive conditioning tasks (Montague et al, 1996; Schultz et al, 1997), in the form of a temporal difference prediction error term.