Development of a learning system for robot control

Abstract

Whilst massive strides have been taken recently in advancing the farming industry with new up-and-coming technologies such as hydroponics and UV farming, the use of automation in these hi-tech farming applications are still relatively novel. As such, existing automated farming applications often incur high costs and have a rigid automation process that offers low customisability. This report therefore aims to explore a low-cost, yet highly versatile way to automate farming. An automated pick-and-place application in the context of hydroponics farming will be developed with the use of a robot manipulator, as well as the YOLO detection system, coupled with a LSTM RNN to perform robust detection and localisation of hydroponic plant pots. The report will also analyse the degree of feasibility in implementing the developed system within autonomous, as well as collaborative forms of application.