Delta and mobile robot

Delta robot picks up and places a fake pepperoni

The GIF shows our robotic system, named R.E.M.Y. (Robot for Expedited Manufacturing of Ya’ pizzas) at work (4x speed).


Roles: Delta Robot Hardware, Mobile Robot Software
Skills: ROS, Python, CAD (SolidWorks), Computer Vision, Hardware
Time: February - June 2019
Team: 9 (MechE, CompSci)

Developed as part of 2.12, Introduction to Robotics at MIT.

The Problem

This project had several goals:

My primary roles were in building the delta robot and developing the software for the mobile robot.

The Process

The first step was to build the delta robot. We were provided with a base design, which we modified for improvement. We redesigned the frame so that it would be more stable and created a fixture for the table so that it would be placed in the same place every time. In addition, we wanted our workspace to be larger than what the base design could reach. To do this, we changed the length of the linkages, used ball joints with a larger angular range, and decreased the size of the end plate.

Calculated workspace for original ball joints (20 degrees) 3D graph showing workspace for delta robot using blue dots.

Calculated workspace for new ball joints (55 degrees) 3D graph showing improved, larger workspace for delta robot using blue dots.

Finally, we redesigned the spacers used to attach the linkages and joints together to reduce friction. The final robot is pictured below. Delta robot with metal frame and linkages surrounding a table with fake pizza within the lab.

For the end effector, we decided on a granular jamming system, which would allow us to relatively easily pick up the various topping shapes and sizes (compared to a motor-actuated gripper). This granular jamming end effector was controlled using a solenoid valve and vacuum pump as shown below. Diagram showing connections between valve, end effector, and limit switches.

The physical device was fabricated using a balloon, coffee grounds, and a 3D printed housing. End effector with balloon, coffee grounds, and 3D printed housing.

Using a camera mounted to the top of the frame, software was developed using OpenCV to detect the colors and shapes of the toppings, as well as the locations in which they should be dropped. Top view of pizza and toppings with green outlines and color labels.

Software was developed for the mobile robot to navigate the boards based on dead reckoning as well as april tags and computer vision for course correction. Finally, the two robots communicated with each other through ROS nodes. The mobile robot would independently reach the table, but only continue on its path to the end once the delta robot was finished placing toppings and had transferred the pizza.

The following is a diagram of the system. Block diagram of robot system software architecture.

The Result

The final system worked on a rough level, but unfortunately, I can’t find a video of the full system or even any videos we took in the later parts of the project, so I guess I can’t provide full proof of that! Some of the challenges were adapting to different lighting to detect the colors of the toppings and ensuring that the failure of the delta robot to perfectly complete the toppings on the pizza would not prevent the robot from at least transferring the pizza and allowing the mobile robot to complete its task. Here are videos of the robot functioning at least in parts during testing (Last GIF is 2x speed):

Delta robot picks up and places a fake pepperoni Delta robot transfers pizza Mobile robot passes waiter obstacle and reaches end

Overall, this was a challenging project with several moving parts that had to all work for success at the system level. As with many projects, we likely would have figured out several of the problems given more time, but it was a good lesson in working with an interdisciplinary team as well as integrating software and hardware! However, we did win an award for the best software in the class! Award for the 2.12 Introduction Robotics 2019 Term Project Competition for Robo's Pizzera Restaurant Automation. Best Software.