The goal of this study is to combine neuronal mechanisms with biomechanics to obtain very fast speed and the on-line learning of circuit parameters. Our controller is built with biologically inspired sensor- and motor-neuron models, including local reflexes and not employing any kind of position or trajectory-tracking control algorithm. Instead, this reflexive controller allows RunBot to exploit its own natural dynamics during critical stages of its walking gait cycle. To our knowledge, this is the first time that dynamic biped walking is achieved using only a pure reflexive controller. In addition, this structure allows using a policy gradient reinforcement learning algorithm to tune the parameters of the reflexive controller in real-time during walking.

Reinforcement learning by direct policy gradient estimation is attractive in theory but in practice leads to notoriously ill-behaved optimization problems. We improve its robustness and speed of convergence with stochastic meta-descent, a gain vector adaptation method that employs fast Hessian-vector products. In our experiments the resulting algorithms outperform previously employed online stochastic, offline conjugate, and natural policy gradient methods.

Reinforcement learning models have long promised to unify computa- tional, psychological and neural accounts of appetitively conditioned be- havior. However, the bulk of data on animal conditioning comes from free-operant experiments measuring how fast animals will work for rein- forcement. Existing reinforcement learning (RL) models are silent about these tasks, because they lack any notion of vigor. They thus fail to ad- dress the simple observation that hungrier animals will work harder for food, as well as stranger facts such as their sometimes greater produc- tivity even when working for irrelevant outcomes such as water. Here, we develop an RL framework for free-operant behavior, suggesting that subjects choose how vigorously to perform selected actions by optimally balancing the costs and benefits of quick responding. Finally, we suggest that tonic levels of dopamine may be involved in the computation linking motivational state to optimal responding, thereby explaining the complex vigor-related ef- fects of pharmacological manipulation of dopamine.

The primary distinguishing feature of temporal-difference (TD) networks (Sutton & Tanner, 2005) is that they permit a general compositional specification of the goals of learning. The goals of learning are thought of as predictive questions being asked by the agent in the learning problem, such as "What will I see if I step forward and look right?" or "If I open the fridge, will I see a bottle of beer?" Seeing a bottle of beer is of course a complicated perceptual act. It might be thought of as obtaining a set of predictions about what would happen if certain reaching and grasping actions were taken, about what would happen if the bottle were opened and turned upside down, and of what the bottle would look like if viewed from various angles. To predict seeing a bottle of beer is thus to make a prediction about a set of other predictions. The target for the overall prediction is a composition in the mathematical sense of the first prediction with each of the other predictions. TD networks are the first framework for representing the goals of predictive learning in a compositional, machine-accessible form. Each node of a TD network represents an individual question--something to be predicted--and has associated with it a value representing an answer to the question--a prediction of that something. The questions are represented by a set of directed links between nodes.

We introduce a class of MPDs which greatly simplify Reinforcement Learning. They have discrete state spaces and continuous control spaces. The controls have the effect of rescaling the transition probabilities of an underlying Markov chain. A control cost penalizing KL divergence between controlled and uncontrolled transition probabilities makes the minimization problem convex, and allows analytical computation of the optimal controls given the optimal value function. An exponential transformation of the optimal value function makes the minimized Bellman equation linear.

Stress and genetic background regulate different aspects of behavioral learning through the action of stress hormones and neuromodulators. They are hypothesized to be related to neuromodulatory levels in the brain. We found that many aspects of animal learning and performance can be described by simple RL models using dynamic control of the meta-parameters. To study the effects of stress and genotype, we carried out 5-hole-box light condition- ing and Morris water maze experiments with C57BL/6 and DBA/2 mouse strains. The animals were exposed to different kinds of stress to evaluate its effects on immediate performance as well as on long-term memory.

Autonomous helicopter flight is widely regarded to be a highly challenging control problem. This paper presents the first successful autonomous completion on a real RC helicopter of the following four aerobatic maneuvers: forward flip and sideways roll at low speed, tail-in funnel, and nose-in funnel. Our experimental results significantly extend the state of the art in autonomous helicopter flight. We used the following approach: First we had a pilot fly the helicopter to help us find a helicopter dynamics model and a reward (cost) function. Then we used a reinforcement learning (optimal control) algorithm to find a controller that is optimized for the resulting model and reward function.

Current road-traffic optimisation practice around the world is a combination of hand tuned policies with a small degree of automatic adaption. Even state-ofthe-art research controllers need good models of the road traffic, which cannot be obtained directly from existing sensors. We use a policy-gradient reinforcement learning approach to directly optimise the traffic signals, mapping currently deployed sensor observations to control signals. Our trained controllers are (theoretically) compatible with the traffic system used in Sydney and many other cities around the world. We apply two policy-gradient methods: (1) the recent natural actor-critic algorithm, and (2) a vanilla policy-gradient algorithm for comparison.

LSTD is O(n2), where n is the number of parameters in the linear function approximator, while iLSTD is O(n). In this paper, we generalize the previous iLSTD algorithm and present three new results: (1) the first convergence proof for an iLSTD algorithm; (2) an extension to incorporate eligibility traces without changing the asymptotic computational complexity; and (3) the first empirical results with an iLSTD algorithm for a problem (mountain car) with feature vectors large enough (n 10, 000) to show substantial computational advantages over LSTD.

Policy gradient methods are reinforcement learning algorithms that adapt a param- eterized policy by following a performance gradient estimate. Conventional pol- icy gradient methods use Monte-Carlo techniques to estimate this gradient. Since Monte Carlo methods tend to have high variance, a large number of samples is required, resulting in slow convergence. In this paper, we propose a Bayesian framework that models the policy gradient as a Gaussian process. This reduces the number of samples needed to obtain accurate gradient estimates.