Project
| # | Title | Team Members | TA | Documents | Sponsor |
|---|---|---|---|---|---|
| 47 | Auto-following Luggage Platform |
David Chen Lyuxing He |
Xiangyuan Zhang | design_document1.pdf design_document2.pdf design_document3.pdf final_paper1.pdf photo1.jpg presentation1.pptx proposal1.pdf video |
|
| # Project Title: Auto-following Luggage Platform Team Members: - Lyuxing He (lyuxing2) - David Chen (sc60) # Problem Sometimes carrying the suitcase might be the most unsatisfactory part of a trip. This dissatisfaction can also grow into annoyance when the amount of luggage is too large to be carried with without the help of transportation tools. Therefore, people have dreamed about a suitcase that can track its owners automatically on its own, without any need of applying external force to steer it. There have been so-called “smart suitcases” made for sale with different features including USB-port for charging, GPS localization, etc. However, the price is too high for the public to afford, and only a few with exceedingly high prices might be capable of achieving the fully automatic following feature. Therefore, we propose the Auto-following Luggage Platform project that aims to solve the problem with much less cost compared to related products available in the market. # Solution We propose a robot platform that autonomously tracks and pursues its owner. It will use a camera as its primary sensor, and an ultrasonic sensor as a fail-safe of that. # Parts needed (we will be providing the funds for purchasing these) - Cuttable metal plate (for building ) - 4 DC motors - Battery - Camera - Rasberry Pie 4 4G - PCB (Motor control) - Ultrasonic sensor We’ll self-supply any additional materials not listed above. # Solution Components ## Robot drivetrain The drive will consist of 2/4 motors. It will take input PWM signals given by Rasberry Pi 4 and drive the motor accordingly. The drivetrain’s left and right motor will be independent, allowing the robot to turn with different speed on left and right. ## Control Algorithm Using the bounding box data calculated from the Human-body identification subsystem, we can calculate the deviation angle, and use PID to track and minimize this error. We will also use a separate algorithm to control the speed of the robot. Using an estimated distance value, we will speed up and slow down the robot accordingly as well. Combining these together and our robot should be able to track the target autonomously. ## Human-body identification subsystem We will use Yolo6 for human recognition and segmentation to produce bounding boxes. Each bounding box will be made into gait silhouettes and used for a gait-matching algorithm to identify the owner of the suitcase. The bounding box of the identified owner will be returned and used to calculate the offset to the camera center, which will be converted to motor signals for the control system to achieve local orientation adjustment. This subsystem will also return a boolean value that represents safety with respect to possible collisions. ## Safety assurance subsystem The ultrasonic sensor equipped will report the distance to obstacles. The robot will stop immediately if the ultrasonic sensor detects a very close object to avoid collisions. # Criterion For Success The machine is able to follow the owner when the owner is in the camera frame, and maintain a safe distance to the owner. The machine is able to locate the owner (put the owner back into the camera frame) autonomously when camera tracking is lost. The machine is able to avoid collisions with obstacles and humans. |
|||||