IROS 2020 - Better Together: Online Probabilistic Clique Change Detection in 3D Landmark-Based Maps

Contributed Conference Paper for IROS 2020 exploring semantic change detection for mapping in a factor-graph SLAM context.
In this paper, we expanded on the work of David Rosen and Fernando Nobre, developing a new method for landmark persistence change detection well suited for the challenges of complicated 3D worlds which wasn’t previously addressed by prior work. Our approach allows for the use of semantic priors on inclusing of features into a single semantic set, allowing for what amounts to joint object persistence estimation in the real world without explicitly tracking any objects.

Undergrad Thesis - An Exploration of Algorithms Enabling a Semantics-Aware Class-Based Probablistic Dynamic SLAM

Undergraduate Thesis by Sam Bateman from the Department of Computer Science in the School of Engineering at University of Colorado - Boulder
This thesis describes the work that I performed persuant to recieving my undergraduate degree at CU Boulder. This thesis was advised by the exceptional Christoffer Heckman. In it, we explore utilizing a semantics aware clique-based persistence filter to handle non-static elements in a environment. Further, multi-agent tracking systems are reviewed and explored to provide semantic and panoptic segmentation of an environment to track persistance statistics of objects in a world online.

Exploration of Numerical Solutions to 1D Schrodinger Equation

1D Time Dependent and Independent Schrodinger Equation Simulations in Jupyter Notebook
Numerical Methods for the Schrodinger Equation Notebook In this project, we explored numerical solutions to the 1D Schrodinger Equation. The vast majority of the content of this project is in the notebook, so I encourage you to check it out. One intresting note, is that the approach for the arbitrary potential of the Time Dependent Schrodinger equation scales to higher dimensions very cleaning by just using the higher dimensional Fourier Transform (seperate integral for each direction).

Sidewalk Following Robot

A Sidewalk Following robot using Deeplab v3 segmentation.
This project is a robot which, for extended distances, can follow and map sidewalk networks in a global, UTM coordinate frame while avoiding bikes, pedestrians and small ground vehicles. Our platform is a Clearpath Jackal, again, utilizing a Nvidia Xavier for perception, 2x Intel Realsense D435 which are frame synced and a Intel NUC for control and planning. We use Deeplab v3 with a Mobile Net backbone trained on Cityscapes for real time segmentation (25-30fps) running with 16bit floating point mode using TensorRT while doing RGB-D ORB_SLAM2 or DSO depending on the sequence for mapping and pose estimates.