User Tools

Site Tools


6a.054.UVA - Reinforcement Learning Approaches to Intent Inference in Autonomous Vehicles

Project - Summary

Our project aims to create a system where a human and a learning agent can interact in a way that allows one to intuitively learn from the other. This system is taking the form of an augmented reality game of air hockey, so as to study how the agent transfers knowledge from a simulation to an actual game and to see how a human reacts to the actions of a learning agent. We have built a version of the game in Unity to be run on a tablet, which will interface with a motion capture lab and a drone running the agent. We have also created a two dimensional Python simulation of the game in which we have tested several reinforcement learning algorithms to determine which is the best candidate to play against the human player. From our tests so far, Deep Q learning has shown the most promising data but it is still unable to consistently beat the AI programmed to play against it in our simulation. We have concluded that while more work is required to train the agent, simulation based training has been very effective in rapidly studying our agents. Furthermore, we believe that the applications of creating a human-agent compatible interface extend over many fields, and with further research we hope to allow humans to have a better understanding of the processes taking place in the machines around them.

Project - Team

Team Member Role Email Phone Number Academic Site/IAB
Peter Beling PI Not Available (434) 982-2066 University of Virginia
Stephen Adams Co-PI Not Available (434) 924-8082 University of Virginia
Maxwell Patek Researcher Not Available Not Available University of Virginia
Kevin Naddoni Researcher Not Available Not Available University of Virginia
Josh Santana Researcher Not Available Not Available University of Virginia
Funded by: Leidos

Project - Deliverables

1 Scenario documentation
2 Inverse reinforcement learning algorithms
3 Autonomous agent demo
4 Report and conclusions

Project - Presentation Video

Project - Documents

FilenameFilesizeLast modified
6a.054.uva_reinforcement_learning_approaches_poster_2017_fall_meeting.pptx504.9 KiB2019/08/14 15:37
6a.054.uva_poster_pdf.pdf562.4 KiB2019/08/14 15:37
6a.054.uva_ppt_presentation.pptx881.4 KiB2019/08/14 15:37
6a.054.uva_poster_ppt.pptx127.0 KiB2019/08/14 15:37
6a.054.uva_final_report.docx4.6 MiB2019/08/14 15:37
6a.054.uva_executive_summary.docx51.3 KiB2019/08/14 15:37
6a.054.uva_confluence_project_page.pdf148.1 KiB2019/08/14 15:37
6a.054.uva_cvdi_mid-year_report.docx236.3 KiB2019/08/14 15:37
projects/year6/6a.054.uva.txt · Last modified: 2021/06/02 17:24 by sally.johnson