Search

Comparative view of the cost function when using the GWO, GD, and PSO algorithms. [Display omitted] • A Policy Iteration Reinforcement Learning-based control approach is proposed. • A Grey Wolf Optimizer algorithm trains the policy neural network. • A comparison with classical Gradient Descent and metaheuristic PSO is offered. • The validation is done on real-time nonlinear servo system position control. • The fast convergence of the suggested approach is proved. This paper presents a new Reinforcement Learning (RL)-based control approach that uses the Policy Iteration (PI) and a metaheuristic Grey Wolf Optimizer (GWO) algorithm to train the Neural Networks (NNs). Due to an efficient tradeoff to exploration and exploitation, the GWO algorithm shows good results in NN training and solving complex optimization problems. The proposed approach is compared to the classical PI RL-based control approach using the Gradient Descent (GD) algorithm, and with the RL-based control approach which uses the metaheuristic Particle Swarm Optimization (PSO) algorithm. The experiments are conducted using a nonlinear servo system laboratory equipment. Each approach evaluated on how well it solves the optimal reference tracking control for an experimental servo system position control system. The policy NNs specific to all three approaches are implemented as state feedback with integrator controllers to remove the steady-state control errors and thus ensure the convergence of the objective function. Because of the random nature of metaheuristic algorithms, the experiments for GWO and PSO algorithms are run multiple times and the results are averaged before the conclusions are presented. The experimental results shows that for the control objective considered in this paper, the GWO algorithm represents a better solution compared to GD and PSO algorithms. [ABSTRACT FROM AUTHOR]

• A combination of Deep Q-Learning algorithm and metaheuristic GSA is offered. • GSA initializes the weights and the biases of the neural networks. • A comparison with classical random, metaheuristic PSO and GWO is carried out. • The validation is done on real-time nonlinear servo system position control. • The drawbacks of randomly initialized neural networks are mitigated. This paper presents a novel Reinforcement Learning (RL)-based control approach that uses a combination of a Deep Q-Learning (DQL) algorithm and a metaheuristic Gravitational Search Algorithm (GSA). The GSA is employed to initialize the weights and the biases of the Neural Network (NN) involved in DQL in order to avoid the instability, which is the main drawback of the traditional randomly initialized NNs. The quality of a particular set of weights and biases is measured at each iteration of the GSA-based initialization using a fitness function aiming to achieve the predefined optimal control or learning objective. The data generated during the RL process is used in training a NN-based controller that will be able to autonomously achieve the optimal reference tracking control objective. The proposed approach is compared with other similar techniques which use different algorithms in the initialization step, namely the traditional random algorithm, the Grey Wolf Optimizer algorithm, and the Particle Swarm Optimization algorithm. The NN-based controllers based on each of these techniques are compared using performance indices specific to optimal control as settling time, rise time, peak time, overshoot, and minimum cost function value. Real-time experiments are conducted in order to validate and test the proposed new approach in the framework of the optimal reference tracking control of a nonlinear position servo system. The experimental results show the superiority of this approach versus the other three competing approaches. [ABSTRACT FROM AUTHOR]