Hello, welcome to Yayun(Daisy) Du’s website. I am a Ph.D. student in Robotics and Control with a minor in solid mechanics at UCLA. Born and raised in a rural village in China, I have been thinking about freeing labor with my parents being farmers. To this end, I chose robotics design and control as my research subject. As an interdisciplinary field, my robotic research involves system design (rigid/flexible PCB board design included), control, artificial intelligence, deep learning, reinforcement learning, and physics-based machine learning-assisted mechanics analysis. With extraordinary leadership, I have been leading several projects since I started my Ph.D. In my doctoral research, I have been working on several projects and collaborating with multiple groups. First, I designed a simple but functional robot with soft straight tails to swim in granular media, and experimental data showed reasonable quantitative agreement with numerical simulation results. Then, using the same robot, I studied its locomotion in fluids and enabled the robot to follow any prescribed 2D trajectory. Meanwhile, I’ve built a low-cost agri-robot with autonomous recharge, navigation, and weed detection (with deep learning) in flax fields, and I have been connecting with plant science researchers from NDSU to test our robot. Based on the agri-robot, I gathered preliminary data for a successful $450k federal grant from the US Dept. of Agriculture and also prepared 33% of an NSF proposal with four PIs ($1.2M) that received ratings of 3 Very Good and 1 Fair (eventually declined). I also collaborated with another NDSU painting group on autonomous robotic painting.
What I am doing Now?
Update 03/18/2021
Designed a compliant roller and implemented passive impedance control with vision that allows robots to realize human-level performance
1). Proposed a framework enabling a robot with only position control to recognize the shape of 3D objects, and plan the coating trajectory autonomously;
2). Used bare minimum ingredients to realize constant-force coating with a compliant roller, in conjunction with a cheap ultrasonic sensor.
1). Designed simple but functional soft flagellated robots to understand the movement mechanism of bacteria in the granular media and fluids;
2). Wrote C++ simulators based on Discrete Elastic Rods (DER) to simulate the robot movement faster than real time.
Design and implement a low-cost (under $500) compact autonomous robot to navigate, localize itself, recognize and spray weeds, and get recharged in narrow spacing crop fields for several weeks:
1). [Design and fabrication]: Design and fabricate the robot incorporating water-jetting the chassis, lathing the axles, 3D printing component holders, and compromising;
2). [Computer vision]: Without accurate expensive GPS, we combined LiDAR, IMU, wheel encoders, and cameras (including visual odometry) to navigate and locate the robot;
3). [Deep learning]: Build our own weed dataset, develope and deploy convolutional neural networks (CNN) on our onboard embedded system, i.e. the robot to enable the robot to differentiate different weeds and spray the corresponding herbicides.
Advisor: Khalid M. Jawed
Dir 1: Simple but functional soft robot inspired by bacteria
Dir 2: Design and control an autonomous weed-control agricultural robot Advisor: Khalid M. Jawed