Biography

Sepehr Saryazdi is a Master’s student in Mathematics at the University of Sydney, having completed an undergraduate degree with majors in Mathematics, Physics and a minor in Data Science. He has worked on several robotics-related problems in the engineering innovation industry, completed several research projects in geometry during his undergraduate degree and is planning to research machine learning innovation from a mathematical perspective. He is currently working as an intern for the SCS Lab under the Engineering Vacation Research program.

Project Title:

Drone Navigation and Precise Landing using Reinforcement Learning

Project Description:

Using drones in delivery services is a fast-growing industry that gained a lot of commercial attention from companies such as Amazon, DHL, and Google. Drone precision landing is critical for the environment’s safety and success of various package delivery missions. We propose an online machine learning-based strategy for navigating the drone to the landing position precisely. In contrast, current methods mostly focus on camera-based precision landing, in which an image target is identified, and an exact landing position is determined. When vision is obscured, these solutions frequently suffer from inaccuracy during severe weather circumstances such as fog and sandstorms. Moreover, a common challenge with traditional autonomous drone navigation using PID is the accumulation of errors in the drone’s knowledge of its location due to internal biases, wind, and other intrinsic or extrinsic factors. We propose using reinforcement learning on a drone physics simulation to train the drone to detect its own drift and thus self-correct its trajectory during navigation or landing. In phase two, we propose deploying the model on a real drone to compare the landing precision to a model-less drone and measure performance.