الفهرس 
INTRODUCTION AND AIM OF THE WORKOnly 14 pages are availabe for public view
 |  | from | 151 |  |  |
| | | from | 151 | | |
Abstract
The research problem of this thesis is centered on agricultural robots. An agricultural robot is used to execute agricultural tasks such as harvesting and picking. The agricultural problem incorporates numerous functionalities. Modern agricultural equipment include automated machinery required for performing specific tasks. These machines greatly benefit from Artificial Intelligence (AI) for their operation.In order for these machines to operate, they require robotic control of the mechanisms which are driven by inverter controlled induction motor drives. Moreover, the efficiency improvement of the operation can be effectuated by using renewable energy technology to supply the required energy to these equipment.This thesis presents the implementation of agricultural robot-related applications using AI methods in order to enhance their performance. Specifically, three applications are considered: Renewable Energy forecasting which can be used to supply the required energy for the robot operation (solar panels) and/or the planning of different agricultural tasks executed by the robot.Temporal Convolutional Network (TCN) is compared to other machine learning methods. The trained model is published as a web service that can be consumed by an agricultural cloud robot through a REST API. The robot can use the predicted data to harness solar energy for the efficient operation of its solar panel. Inverse kinematics which is used to control the robot manipulators. Extensive survey and analysis is to be conducted for the IK problem of redundant manipulators formulated as a Quadratic Programming (QP) optimization problem solved by different kinds of recurrent neural networks (RNNs). A comparison between different RNN QP solvers is to be performed. Motor Control which is necessary for the robot movement. An extension of the three-level Space Vector Pulse Width Modulation (SVPWM) algorithm implemented using simple feed-forward ANNs and employed inside a scalar control drive system that has been previously presented is to be proposed for the n-level case. The implementation steps of the four-level case are to be presented along with simulation results to demonstrate the validity of the proposed scheme.