E D U B O T S
SHAPING THE FUTURE OF ROBOTICS
Modification Of The Dash Robot, by Miles Crabill
In the 2015 spring semester, I was part of a team of students working with Dr. Kellar Autumn, Director of the Center for Entrepreneurship, adding features to Dash Robotics’ Dash Beta. Fellow computer science majors, Quinn Rohlf, Jeff Shaw, Ben Whitenack, and I wanted to work on automated robots, but we knew we had to start small. Dr.
Autumn introduced us to UC Berkeley’s Dash Beta, a small form factor six-legged robot modeled after the cockroach. Its body is laser-cut carbon fiber and cardboard, with a custom-built transmission, a battery, and “DashBoard,” which is an Arduino mounted on a custom chip. But don’t let that intimidate you—it’s small, cheap, and easy to assemble.
As an added bonus, the Dash’s firmware and related software, including the iOS app to control it, are all open source. While working with Dash, we realized that the built-in “Bump” function, which is sup- posed to make Dash react and move away from objects when it detects an imminent collision, doesn’t work very well. It uses the Dash’s infra- red sensors to detect obstacles, but we found that the function had been tuned to specific lighting environments and that the IR sensors only worked at certain angles. We decided to add a flex sensor to the front of Dash to detect collisions.
At first, we had hoped to add two sensors, imitating cockroach antennae, but we were limited by the number of GPIO pins on the DashBoard and didn’t want to configure a breakout board for the prototype. Through a combination of readings from the built-in infrared sensor and our flex sensor we created a more robust collision detection function for Dash. We added a wide strip of cardboard to the flex sensor, serving as a bumper that can help detect collisions from a variety of angles. Still, we could just detect collisions, but not calculate the angle at which the collision occurred; instead, we just assumed it was from the front.
From ideation to purchasing parts to building the final flex sensor battering ram, our team had about five weeks of working on and off before the project deadline. As a result, that was as far as we got—but there’s plenty of room for future work. One possibility would be to have two separate flex sensors, which would allow a calculation of the angle at which Dash is colliding with objects. This would enable a totally improved set of maneuverability responses for Dash. After completing our project, we got in touch with the PhD students at UC Berkeley who designed the Dash. They were really excited to see people expanding Dash’s functionality and we were able to make some suggestions for post-beta development of Dash. Some of the code we wrote is available here: https://github.com/milescrabill/dashmod.
Lewis & Clark College, www.lclark.edu