Embedded artificial intelligent (ai) To navigate cart follower
Loading...
Date
2018-06
Authors
Tang, Khai Luen
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The concern of the societies in creating a quality life for everyone without laying aside
of the right of disable person leads to research on designing and fabricating autonomous
robot. Wheelchair user usually faces the problem of carrying luggage along during travel
as they need both of their hands to navigate their wheelchair. One of the solution for the
problem is to create an Artificial Intelligent (AI) cart follower. Therefore, this research is
to create an AI system for the AI cart follower with a visual based sensor. The visual
based sensor gathered the information of the width, height, angle, x and y coordination of
the colour pattern board which situated behind the wheelchair and translate this
information into relative position information which enable the cart to follow the
wheelchair. This translation can be done in neural network. However, the data needs to
be collected in such a way that the output distance is manipulated between 20cm to 69cm
and the output angle is manipulated between -30 to 30 with its restriction for each case.
The test MSE value is used to evaluate the performance of NN and validation MSE value
is used to prevent overfitting. The weights and biases generated through the training
process is depended on the training algorithm, initial weights and biases for training and
the dataset used in the process. The training algorithm may also vary with different sets
of parameters, number of neurons and activation function. The set of parameters used in
traingd are lr, max_fail, min_grad, goal, time, and epochs. The final weights and biases
generated with the minimum MSE performance after several run is used to train NN in
the FPGA together with the structure of NN obtained in Simulink. The implementation of
neural network on the FPGA can be done through software or hardware configuration.
However, the floating-point operation circuit needs to be built to ensure the NN on FPGA
is functioning.