A generic architecture for adaptive agents based on reinforcement learning

Self-organizing state aggregation for architecture design of Q-learning

“…A typical mathematical model of a cart-pole system can be represented as Eqs. (12) and (13). Also, the cart-pole model by a personal computer is simulated using the fourth order Runge-Kutta [3].…”

“…By taking action 'a' based on state 's', the algorithm evaluates the feedback rewards and updates its Q values from the rewards sequentially. In an acceptable learning period, the Q-learning algorithm performs excellently, as evidenced by its successful application to collision-avoidance, homing, and robot cooperation [4,6,7,9,13].…”

A novel approach for multi-agent-based Intelligent Manufacturing System

“…The theory of multi-agent system and distributed artificial intelligence (DAI) 0020 provides feasible technical support to modeling and realization of Intelligent Manufacturing System, which becomes one of the research hotspots in the manufacturing domain [27]. Manufacturing process is a typical multi-agent question solution process, and every department (or segment) in manufacturing system is equal to an Agent in the process.…”

A fuzzy Actor–Critic reinforcement learning network

“…Reinforcement learning methods have been widely applied to a variety of problems, such as autonomous robot navigation and nonlinear control [7,22]. However, most conventional reinforcement learning methods use a discrete state space to describe the interaction between the agent and its environment.…”