river logo.png


"Deep Reinforcement Learning based Robot Navigation in Dynamic Environments using Occupancy Values of Motion Primitives" authored by Neşet Unver Akmandor, Hongyu Li, Gary M. Lvov, Eric Dusel and Taşkın Padir has been accepted to the Proceedings of the 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2022)!

On the code base, I mostly worked on generalizing the roslaunch files/reworking the file structure of the framework so that adding and supporting new robots was a breeze. I wrote the majority of the Simulation section of the README to describe on how to run the framework with the reworked launch files. I wrote the initial version of the bash script to install the framework. I also started to develop an imitation learning branch of the repo that I hope I can disclose more about in the future.

On the paper itself, I created visualizations of the success/failure of different local planners to easily compare and contrast our method to different state of the art methods. Other than the visualizations, I primarily revised the grammar and wording to increase clarity for the entire paper.