scholarly journals Performance Analysis of Deep Neural Network Controller for Autonomous Driving Learning from a Nonlinear Model Predictive Control Method

Electronics ◽  
2021 ◽  
Vol 10 (7) ◽  
pp. 767
Author(s):  
Taekgyu Lee ◽  
Yeonsik Kang
Keyword(s):  
Neural Network ◽  
Model Predictive Control ◽  
Predictive Control ◽  
Nonlinear Model ◽  
Deep Neural Network ◽  
Nonlinear Model Predictive Control ◽  
Control Technique ◽  
Computational Time ◽  
Target System ◽  
Neural Network Controller

Nonlinear model predictive control (NMPC) is based on a numerical optimization method considering the target system dynamics as constraints. This optimization process requires large amount of computation power and the computation time is often unpredictable which may cause the control update rate to overrun. Therefore, the performance must be carefully balanced against the computational time. To solve the computation problem, we propose a data-based control technique based on a deep neural network (DNN). The DNN is trained with closed-loop driving data of an NMPC. The proposed "DNN control technique based on NMPC driving data" achieves control characteristics comparable to those of a well-tuned NMPC within a reasonable computation period, which is verified with an experimental scaled-car platform and realistic numerical simulations.

Aero-engine direct thrust control with nonlinear model predictive control based on linearized deep neural network predictor

2019 ◽  
Vol 234 (3) ◽  
pp. 330-337 ◽  
Author(s):  
Qiangang Zheng ◽  
Yong Wang ◽  
Fengyong Sun ◽  
Haibo Zhang
Keyword(s):  
Neural Network ◽  
Model Predictive Control ◽  
Predictive Control ◽  
Nonlinear Model ◽  
Deep Neural Network ◽  
Sliding Window ◽  
Nonlinear Model Predictive Control ◽  
Thrust Control ◽  
Direct Thrust Control ◽  
Linearized Model

A novel nonlinear model predictive control method for aero-engine direct thrust control is proposed to improve engine response ability and reduce computational complexity of nonlinear model predictive control. The control objective of the proposed method is the thrust directly instead of the measurable parameters. The linearized model based on online sliding window deep neural network is proposed as predictive model. The online sliding window deep neural network has strong fitting capacity for nonlinear object and adopted to fitting the transient process of engine. The back propagation is adopted to obtain linearized model of online sliding window deep neural network, which greatly reduce the calculated amount. The comparison simulations of the popular nonlinear model predictive control based on extended Kalman filter and the proposed one are carried out. The simulation results show that compared with the popular nonlinear model predictive control, the proposed nonlinear model predictive control not only has the better response ability but also has reduced computational complexity greatly, nearly reduce computation time more than 35 ms.

scholarly journals LVD-NMPC: A learning-based vision dynamics approach to nonlinear model predictive control for autonomous vehicles

2021 ◽  
Vol 18 (3) ◽  
pp. 172988142110195
Author(s):  
Sorin Grigorescu ◽  
Cosmin Ginerica ◽  
Mihai Zaha ◽  
Gigel Macesanu ◽  
Bogdan Trasnea
Keyword(s):  
Neural Network ◽  
Model Predictive Control ◽  
Predictive Control ◽  
Nonlinear Model ◽  
Autonomous Vehicles ◽  
Process Model ◽  
Network Performance ◽  
Vision System ◽  
Nonlinear Model Predictive Control ◽  
Scaled Model

In this article, we introduce a learning-based vision dynamics approach to nonlinear model predictive control (NMPC) for autonomous vehicles, coined learning-based vision dynamics (LVD) NMPC. LVD-NMPC uses an a-priori process model and a learned vision dynamics model used to calculate the dynamics of the driving scene, the controlled system’s desired state trajectory, and the weighting gains of the quadratic cost function optimized by a constrained predictive controller. The vision system is defined as a deep neural network designed to estimate the dynamics of the image scene. The input is based on historic sequences of sensory observations and vehicle states, integrated by an augmented memory component. Deep Q-learning is used to train the deep network, which once trained can also be used to calculate the desired trajectory of the vehicle. We evaluate LVD-NMPC against a baseline dynamic window approach (DWA) path planning executed using standard NMPC and against the PilotNet neural network. Performance is measured in our simulation environment GridSim, on a real-world 1:8 scaled model car as well as on a real size autonomous test vehicle and the nuScenes computer vision dataset.

Sign in / Sign up

Export Citation Format

Share Document