TurtleCups

Introduction

The goal of this project is to explore the difficulties in L5 autonomous driving, specifically programming the turtlebot to navigate between a known starting and finish point by leveraging the use of camera data to navigate around obstacles.

​

The sensor architecture is similar to ones being deployed by major self-driving car companies at the time of writing this and the turtlebot project can help illustrate the inherent difficulties of such a setup with an attempt to provide resolution approaches. Major problems we’re going to attempt to solve are image segmentation (to provide the start, end and obstacle coordinates to the robot), path planning (utilizing RRT) and turtlebot controls.

​

Why this problem? L5 has been a big topic in the tech industry for a while now, with several companies trying to tackle the problem (e.g. Waymo, Cruise..). It will enable safer roads and also give people better utilization of their time. Additionally, robots are being used in places such as Amazon warehouses where automation will significantly reduce mundane human tasks. This last use case is similar to our project setup where one could have a warehouse with a central camera to enable localization and path planning with robots being orchestrated around with onboard controls algorithms, think amazon KIVA.