We were inspired by the recent uptick in AI, which made us think what if we made something that rolled around campus and made people's day by spreading positivity and sharing helpful resources around campus.
What it does
The robot uses simple machine learning algorithms to interpret and provide auditory responses to voice inputs. Certain information about resources are stored on the bot's "brain" that can be relayed to the user standing in front of the bot's camera view. Alongside the computer vision and machine learning, we implemented an autonomous mode which if the robot hits an obstacle it will stop and move in a different direction.
How we built it
We built it by hardware hacking into a Double telepresence robot, which allowed us to use some of the onboard functions to control the robot automatically. The ML and CV models were prebuilt and came from Apple's database, specially using SqueezeNet and CoreML frameworks/SDKS. Through the use of xCode and Objective C we built an app that interfaces with the bot and allows for the autonomous controls to be sent to and from the tablet device.
Challenges we ran into
We ran into numerous challenges through implementation in xCode, hardware hacking the robot, understanding how the robot's code and wireless communication worked, learning how to implement CV and ML models from Apple, and much more.
Accomplishments that we're proud of
Getting over most if not all the challenges listed above and just taking the shot at trying to bring such an ambitious idea to life.
What we learned
How to program and use Objective C, hardware hacking, programming autonomous modes on a dual drive robot, implementing machine learning models, implementing computer vision on a tablet, and working together in a collaborative and efficient team.
What's next for Double Down Robotics
We're looking to improve on the UI of the app interface and make the path finding and computer vision algorithms much faster. As of now the path finding and calculations take a good amount of the time so the robot moves at a slow pace and seems to stutter and wait intermittently when it moves.