Publication:
Embedded system on robot: fpga hardware implementation of neural networks algorithm for mobile robot autonomous navigation

Loading...
Thumbnail Image
Date
2008-05-01
Authors
Hoe, Lay Fang
Journal Title
Journal ISSN
Volume Title
Publisher
Research Projects
Organizational Units
Journal Issue
Abstract
This project is attempted to demonstrate the FPGA (Field Programmable Gate Array) hardware implementation of neural network (NN) algorithm for mobile robot autonomous navigation. The NN topology was designed with adequate NN algorithm and trained using NN toolbox in Matlab®. The sensory input system of the mobile robot was designed to create the training data pairs for the NN training. The collision-free path finding of the obstacle avoidance task was considered for the training data pairs design. An optimal NN topology was determined after having a well-trained with good performance. The optimal NN topology design was tested with the testing data pairs to determine its stability and reliability. The final weight was taken from the last train of the optimal NN algorithm design for the FPGA hardware implementation. The embedded system design was consisted of two parts which are hardware system design and the NN algorithm software programming. The hardware system was generated using SOPC (System-on-a-programmable-chip) Builder of Quartus ® II design software and downloaded to the FPGA. The software design was developed using Nios® II IDE (Integrated Design Environment) and integrated with the hardware system. The simulation and testing corresponding to different obstacles situations was done to ensure the functionality, accuracy and reliability of the FPGA hardware implementation of the neural controller design. The result shows that the FPGA hardware implementation of neural controller with NN algorithm was well-established.
Description
Keywords
Citation