Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Videos

Training recurrent neurocontrollers for real-time applications.

Danil V Prokhorov¹

¹Toyota Technical Center, Division of Toyota Motor Engineering and Manufacturing North America (TEMA), Ann Arbor, MI 48105, USA. dvprokhorov@gmail.com

IEEE Transactions on Neural Networks

|August 3, 2007

Summary

This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by

Same author

Model-Free Dual Heuristic Dynamic Programming.

IEEE transactions on neural networks and learning systems·2015

Same author

GrDHP: a general utility function representation for dual heuristic dynamic programming.

IEEE transactions on neural networks and learning systems·2014

Same author

Toyota Prius HEV neurocontrol and diagnostics.

Neural networks : the official journal of the International Neural Network Society·2008

Same author

Special issue on neural networks for feedback control systems.

IEEE transactions on neural networks·2007

Same author

Training winner-take-all simultaneous recurrent neural networks.

IEEE transactions on neural networks·2007

Same author

Training recurrent neurocontrollers for robustness with derivative-free Kalman filter.

IEEE transactions on neural networks·2006

Same journal

Universal perceptron and DNA-like learning algorithm for binary neural networks: LSBF and PBF implementations.

IEEE transactions on neural networks·2013

Same journal

Guest editorial: special section on white box nonlinear prediction models.

IEEE transactions on neural networks·2011

Same journal

Data-based fault-tolerant control of high-speed trains with traction/braking notch nonlinearities and actuator failures.

IEEE transactions on neural networks·2011

Same journal

Guest editorial: special section on data-based control, modeling, and optimization.

IEEE transactions on neural networks·2011

Same journal

Neural network-based multiple robot simultaneous localization and mapping.

IEEE transactions on neural networks·2011

Same journal

Data-driven model-free adaptive control for a class of MIMO nonlinear discrete-time systems.

IEEE transactions on neural networks·2011

See all related articles

This study presents a novel method for training recurrent neurocontrollers, enhancing real-time control through adaptive internal states and robust initial training. The approach improves performance in critical applications like vehicle systems.

Area of Science:

Control Systems Engineering
Artificial Intelligence
Robotics

Background:

Recurrent neurocontrollers are crucial for real-time applications but require robust training methods.
Existing training techniques often struggle with high-fidelity models and embedded deployment.

Purpose of the Study:

To introduce a new, enhanced approach for training recurrent neurocontrollers suitable for real-time embedded systems.
To improve the robustness and adaptability of neurocontrollers through a two-stage training process.

Main Methods:

Initial training utilizes an enhanced derivative-free Kalman filter method on high-fidelity models.
Real-time adaptation employs simultaneous perturbation stochastic approximation (SPSA) and an adaptive critic (RNN trained by stochastic meta-descent).

Related Experiment Videos

The neurocontroller's internal state (short-term memory) is adapted in real-time, while weights remain fixed post-initial training.

Main Results:

The proposed method successfully trains recurrent neurocontrollers for real-time applications.
Application to electronic throttle control and hybrid electric vehicle control demonstrated apparent performance improvements.
The combination of SPSA and adaptive critic, along with SMD for critic training, proved effective for real-time adaptation.

Conclusions:

The novel two-stage training approach enhances recurrent neurocontroller performance in real-time embedded systems.
This method offers a promising direction for developing more robust and adaptive control systems.
The successful application in automotive control highlights the practical viability of the technique.