Survey Of Reinforcement Learning Research Articles

A tutorial introduction to reinforcement learning

In this paper, we present a brief survey of reinforcement learning, with particular emphasis on stochastic approximation (SA) as a unifying theme. The scope of the paper includes Markov reward processes, Markov decision processes, SA algorithms, and widely used algorithms such as temporal difference learning and Q-learning.

Survey of Reinforcement Learning based on Human Prior Knowledge

At present, task planning is mainly solved by rules or operational research methods. The time complexity and space complexity of these methods increase exponentially with the growth of problem size. Thus, it is challenging to solve large-scale problems. Moreover, it is helpless to solve dynamic task planning problems. Reinforcement learning (RL) is often used to solve the dynamic planning problem of continuous decision making. RL agent constantly interacts with the environment to achieve the expected goal. However, the RL method meets challenges when handling the problem with ample state space. More sampling and exploration are needed to update the strategy gradient, and the convergence speed is slow. Humans ensure a quick start of learning by using prior knowledge, which reduces the exploration time of problems. Thus, we review the methods which combine human prior knowledge with RL through temporal node classification of human prior knowledge combined with RL. In this way, agents effectively reduce the sampling and exploration of the environment and eventually get the optimal strategy faster. Finally, we propose the development direction.

Reinforcement Learning with Temporal Logic Constraints

Abstract Reinforcement learning (RL) is an agent based AI learning method, where learning and optimization are combined. Dynamic programming is then performed iteratively, based on reward and next state observations from the system to be controlled. A brief survey of RL is given, followed by an evaluation of a recently proposed method to include temporal logic safety and liveness guarantees in RL, here combined with classical performance optimization. RL is based on Markov decision processes (MDPs), and to reduce the number of observations from the system, a modular MDP framework is proposed. In the learning process, it is then assumed that some parts of the system are represented by known MDP models, while other parts can be estimated by observations from the real system. Local information from the modular system may then be used to reduce the computational complexity, especially in the handling of safety properties.