2019 30.007 project

30.007 is a mod that every SUTD Engineering Product Development (EPD) student is required to take. The mod is a giant project fest. Students are to form groups independently and, in ~13 weeks, would ideate, research, and prototype a product according to a theme given (if I remember correctly, my year was intelligent robotics). On top of all of these, there are still presentations and reports that need to be submitted regularly. It is a mod infamous for giving students headaches and sleepless nights.

Group members and dynamics

There are 6 of us in this project, a slightly bigger group as the average was 5. We divided the roles as follow:

1. Me (Electrical + Software)
2. Electrical
3. Electrical + Mechanical
4. Mechanical + Project management
5. Mechanical
6. Mechanical

I would say that we worked rather well together, and we had clear, distinct roles. This is my first time tackling such a massive project. I remember being exposed to many concepts/tools such as Raspberry Pi and using OpenCV in python. My role in this project is general Arduino programming and chunking out a follow-me system.

Project overview

• We found out that the number of middle-income & elderly travellers has been rising in recent years.
• Aiport experience plays a crucial role in making travelling a pleasant experience for this new demographic.
• And one way to improve on it is by making moving luggage more convenient around the airport.
• Thus our smart trolley project was born; we name it Jack Me cause we used a car jack as a mechanism for the trolley to move its platform up and down.

Functionalities

We want to achieve the following:

1. A motor-powered trolley
2. An unfolding ramp that allows one to slide baggage up and down it
3. A trolley platform that can move up and down to facilitate the moving of luggage from to/from the car trunk
4. Some sort of follow-me and remote control functionality
5. Ultrasonic obstacle sensor
6. A simple interactive touch screen interface

This was quite a lot to do in 14 weeks, considering we still had lessons and this is our first big project. Nevertheless, we managed to achieve our goals w.r.t 1 and had a proof of concept for 2, but 3, 4, 5.

Technical notes

The trolley itself

The trolley is custom-built from aluminum extrusions and welded together with metal sheets (way to go mechanical team!). As mentioned earlier, there is a car jack on the trolley that can raise the platform up or down. It is also motor-powered with a side ramp with rollers to allow the sliding of baggage to and from a car trunk

Sensor and peripherals

• A simple LCD touch screen hooked up to the Raspberry pi3 running a GUI made from python with Tkinter
• An Arduino mega
• An IR sensor to check how tall the platform has gone up.
• There are ~5 (?) HC-SR04 ultrasonic sensors on the trolley. 3 at the handle and 2 at the sides
• A raspberry pi camera hooked up to the Raspberry Pi3
• Serial Bluetooth Module HC-05 to listen to inputs from a smartphone phone

The brains

• The raspberry pi runs an OpenCV and Tkinter GUI python script, and it is Serially connected to the Arduino mega board
• The OpenCV script implements a circular purple image object detection
• It will try to “follow it” constantly trying to center to the purple circular object on the camera image stream
• This is done by sending Serial data to the Arduino mega to move the trolley hence the mounted camera left or right
• The blue tooth module also sends Serial data to the Arduino mega
• The Serial connection is essential as the Arduino mega is always listening to character inputs from either the phone or rpi3 to update the trolley state
• For example, Serial character ‘U,’ ‘D’ indicates to switch trolley state to levelling the platform jack
• ‘F,’ ‘L,’ ‘R’ is for the trolley to switch state to move forward, left, right, and ‘z’ and ‘p’ is to indicate whether openCV is activated or not
• The rest of the IR and Ultrasonic sensor is also hooked up to the Arduino mega

Afterthoughts

Even though we managed to develop a prototype, we bit more than we can chew, alot of the functionalities aren’t super robust and serves more as proof of concept than actual use. Initially, the follow-me system is ultrasonic based. The target will carry an ultrasound beacon that will keep on emitting ultrasonic waves to 2 receivers (from there, we would triangulate the position). However, I realize it was a bit too technically challenging to implement so I pivoted to computer vision which was pretty hot back then. I thought we could strap a Rpi3 into it with a pi camera and perhaps do some sort of simple object recognition then follow that object. Even though this could also be considered a technically difficult challenge to implement, there was more documentation and guides on this subject. All I need to do is sit down, research, and learn how to do it. I think that was one of my biggest take away from this project. I don’t even know what computer vision is at all but thank the lord for google, and somehow, with a lot of grit, I managed to pull through.

Sometime in the future I will look for a way to organize or show these pictures better. For now this should suffice.