The Design and Control of an Economical Autonomous Surface Vehicle for Object Detection

Abstract

Autonomous underwater vehicles have paved the way for intelligent agents to assist with critical marine tasks such as retrieval, inspection, research and relief without the need for human intervention. While the applications of these systems are far ranging, these vehicles are usually very expensive, therefore limiting the applications for non-well funded problem spaces. This project involved fabricating a low-cost autonomous surface vehicle capable of detecting and navigating towards tagged objects in a body of water for implementation in foreign object retrieval applications such as trash pickup. The vehicle performs all control autonomously, using only on-board processors, without the need to communicate with an external computer. Control is implemented first using a PID controller for proof-of-concept, and then a candidate LQR controller and observer are presented for more optimal control. We calculated the dynamics model for the three degree-of-freedom system from the derived equations of motion in order to design this LQR controller. The vehicle design was developed upon work done by previous students and post-doctorates in the Leonard Lab at Princeton University. Due to the COVID-19 pandemic, all work for this project had to be moved to an at-home environment. As such, the objective of the project changed from executing our original design to the creation of a proof-of-concept prototype, which is the vehicle we have described above. The original design for the surface vehicle can detect objects at the bottom of a larger tank, navigate towards those objects, and retrieve them with an attached gripper. This report includes both a description of the prototype vehicle created as well as our original design for a vehicle, which we call SWAN.