Development of water surface robot system for lake sanitation and sampling / Ahmed Abdullah Omar

Ahmed Abdullah , Omar (2022) Development of water surface robot system for lake sanitation and sampling / Ahmed Abdullah Omar. Masters thesis, Universiti Malaya.

	PDF (The Candidate's Agreement) Restricted to Repository staff only Download (202Kb)
	PDF (Thesis M.A) Download (4004Kb)

Abstract

There are many exciting individual robot designs and strategies in either water sanitation or sampling tasks. However, rarely a robot design considers handling both tasks because they do not share similar sensing technology or similar data processing framework. Also, low-cost robot designs have become limited to onboard computational resources; sensor data is captured locally, and the mechanical aspects of the robots are mainly pre-programmed or remotely controlled. In low-cost robots, automation is hardly ever considered. This dissertation explores a design where a low-cost water surface robot can tackle sanitation and sampling tasks. Central to the design is real-time access to mechanical control and sensor data for off-board processing and analytics. Off-board computation allows other researchers to deploy advanced algorithms without restrictions on computing resources. Only the results need to be communicated back onto the robot for real-time actions. The dissertation proposes three objectives; to design and develop the water surface robot structure, control and electronics, design and develop the APIs for remote robot connection, data communication and positioning, and evaluate the robot’s performance through case studies. Regarding results, a low-cost robot system that can glide on the lake surface has been developed. The robot system showcases PVC pipes for the body with a load up to 17kg. The electronic design covers the cooling system, battery level monitoring, voltage control and sensor reading. The software design features a mobile app for remote access, live video streaming, mapping system and communication APIs. The robot networking includes an RTK server setup for accurate localization and positioning. Two case studies test the robotic system performance. Case study 1 tests the robot’s APIs in implementing a genetic algorithm for wayfinding with RTK-based localization. Case study 2 tests the robot’s APIs in training a deep learning model for garbage detection based on the robot's live stream. The dissertation shows that an open robotic design can coordinate contributions by various parties on a single platform; robotics engineers can focus on improving the robot hardware and controls towards enabling live data. AI researchers can test automation theories on robot data without the need for robot hardware and control know-how. Such design allows anyone to drive the robot over the internet without the physical need to be on site. Remote driving can improve lake sanitation and sampling activities and can significantly promote citizen science.

Actions (For repository staff only : Login required)