Xuexuan Peng

Xuexuan Peng

Hi, I am an MEng student in Mechanical Engineering at Duke University. I worked with Simona Tomar during this Capstone class (ME 555) with advice from Prof. George Delagrammatikas. Here is my LinkedIn and feel free to connect! 

The project aims to develop a wheeled robot to automatically avoid obstacles on its path. The robot should identify and record the obstacle in its database. A LiDAR sensor is used to detect and map the obstacles in task 1 and a line tracking sensor helps the robot follow routes in task 2.

The project can be divided into two tasks for a team of two people: optimal structure design and ROS coding. I performed the first task, which includes designing chassis structure, planning hardware layout, and assembling hardware. In addition, I also set up the jetson coding environment such as flashing an SD card, registering WIFI, and installing startup packages.

Why does chassis design cover such a big part, and why not use the Jetbot structure directly? – Our robot is unique in using LiDAR sensor and Nvidia Jetson nano microcontroller, with electronic components from an ELEGOO robot kit. There is no chassis online that can accommodate such a combination of components.

Starting from the jetbot structure, I envisioned, 3D printed, and tested different shapes of chassis to fit all the components. The chassis consists of four layers: the bottom layer for the motors and wheels; the first layer for the microcontroller, power supplies, and motor driver; the second layer for the microcontroller and ultrasonic sensor; the third layer for the LiDAR sensor. The LiDAR sensor has to be at the top to be able to collect effective data about surrounding obstacles. Here is a tutorial about the components and how they connect with each other. I also wrote nearly all of the content for the project website.

The prototyping process may take a long time for students without prior experience. I therefore made a tutorial of how I prototyped the robot chassis from nothing. One KEY lesson I learnt is to simulate the environment in CAD software first to locate the fixtures. This way saves much time than to design a structure, 3D-print it out, test, and redesign. There are many online, free models that can be directly imported to CAD software and you only need to design the enclosure/fixtures based on them.

I also made two tutorial websites for students to learn 3D printing and laser cutting quickly. I used laser cutting to determine the general layout. Laser cutting costs much less time than 3D printing and is useful for initial prototyping. But 3D printed structure gives more detailed information as to what should be improved next. More details about these two methods, including tips, are in the tutorial websites.