Large Scale Mobile Robotic 3D Printing Project

While useful, traditional gantry-based 3D printers have a few significant disadvantages. The printing scale is strictly confined to the printer platform, and the printing process has relatively low efficiency. Using mobile robots for 3D printing can be one potential solution. However, highly accurate localization of these robots is the key challenge to this new possibility.

Accurate 3D printing requires millimeter level accuracy of the printhead, but traditional cost-effective localization methods are at most of centimeterlevel accuracy. Our research will use deep neural networks to fuse laser sensors and cameras, leading to a new method of localization suitable for mobile robotic 3D printing. Such a method will also be able to incorporate other sensors such as odometry from IMU/wheel-encoders, to further increase the accuracy and efficiency of localization.

We worked on both the hardware and software platform for testing and improving the localization accuracy and was able to demonstrate a large-scale 2D printing (at least 3m by 3m size) by the end of the project.