Knowledge Implementation and Transfer With an Adaptive Learning Network for Real-Time Power Management of the Plug-in Hybrid Vehicle

Abstract

Essential decision-making tasks such as power management in future vehicles will benefit from the development of artificial intelligence technology for safe and energy-efficient operations. To develop the technique of using neural network and deep learning in energy management of the plug-in hybrid vehicle and evaluate its advantage, this article proposes a new adaptive learning network that incorporates a deep deterministic policy gradient (DDPG) network with an adaptive neuro-fuzzy inference system (ANFIS) network. First, the ANFIS network is built using a new global K-fold fuzzy learning (GKFL) method for real-time implementation of the offline dynamic programming result. Then, the DDPG network is developed to regulate the input of the ANFIS network with the real-world reinforcement signal. The ANFIS and DDPG networks are integrated to maximize the control utility (CU), which is a function of the vehicle's energy efficiency and the battery state-of-charge. Experimental studies are conducted to testify the performance and robustness of the DDPG-ANFIS network. It has shown that the studied vehicle with the DDPG-ANFIS network achieves 8% higher CU than using the MATLAB ANFIS toolbox on the studied vehicle. In five simulated real-world driving conditions, the DDPG-ANFIS network increased the maximum mean CU value by 138% over the ANFIS-only network and 5% over the DDPG-only network.

The Intelligent Path Planning System of Agricultural Robot via Reinforcement Learning

“…In the study of Beomjoon et al, they utilize a feature extraction module to extract local features, and then use inverse reinforcement learning to learn a cost function from human experience to assist local path planning, and make the agent trajectory similar to human behavior to improve the safety of the robot [30]. Zhou et al combine DDPG with ANFIS network to realize real-time dynamic planning of vehicles to maximize vehicle energy efficiency and state of charge [31]. Lin et al combine DDPG with LSTM network to fully memorize and utilize the robot's past state in order to complete the collision-free path planning and picking of tasks of picking robots in the orchard [32].…”

Development of a Method for Evaluating the Technical Condition of a Car’s Hybrid Powertrain

“…During the operation of the TB, an imbalance of the elements, in terms of capacity, internal resistance, and other parameters, appears and reduces the efficiency of the battery as a whole [30]. The car reacts to the reduction of the TB quality by increasing the fuel consumption, an incorrect indication, and, generally, by decreasing the power.…”

“…At the same time, there are methods that make it possible to successfully identify dependencies of the complex functions based on the experimental studies with the limited data. Among such approaches, one can single out the use of fuzzy inference systems [30]. The advantages of the latter include the possibility of formalizing and using primary information about the phenomena of the study.…”

“…The choice of the Gaussian function relates with the dependence of the internal combustion engine and the traction battery. The advantage of this membership function is that only two parameters are needed for its task: b and c. The method of two stage identification of the nonlinear Rothstein dependence [30] was used to synthesize a fuzzy identification system for the dependence P k = f (I, U, S, V c ). According to this method, firstly, the base of fuzzy rules of the form "if something" (structural identification) is formed, then, a parametric identification of the dependence is performed by finding such weights of rules and parameters of membership functions of fuzzy terms that minimize deviations of the results of fuzzy modelling from the experimental data.…”

Pedestrian-Aware Supervisory Control System Interactive Optimization of Connected Hybrid Electric Vehicles via Fuzzy Adaptive Cost Map and Bees Algorithm

“…For PHEVs, chargedepleting (CD) mode and charge-sustaining (CS) mode [8] are widely used that enable to maximize the usage of electricity at sufficient charge and maintain a safe state of charge at insufficient charge [9]. To reduce the development workload for energy management controllers, Quan et al research a transferable representation modelling routine, where two artificial intelligence technologies of deep neural network [10] and Gaussian process regression [11] are developed to cooperate with an adaptive neuro-fuzzy inference system for knowledge transfer of the energy management controller. To maximize effectiveness of heuristic control strategies, global optimization approaches have been applied to optimize their parameters.…”