Abstract
The University of Central Florida invention is a system and method for autonomous vehicle (AV) navigation. To enhance decision-making and safety, the system applies deep reinforcement learning that enables high-level policy creation for safe, tactical decision-making. It also optimizes social utility and increases sample efficiency and safety, making significant strides in autonomous vehicle operation.
Today’s roads support a mix of autonomous and human-driven vehicles (HVs), which must learn to co-exist by sharing the same road infrastructure. To attain socially desirable behaviors, autonomous vehicles (AVs) must be instructed to consider the utility of other vehicles around them in their decision-making process. Yet, despite the advances in the autonomous driving domain, AVs are still inefficient and limited in terms of cooperating with each other or coordinating with vehicles operated by humans. The UCF invention offers a solution with a system and method that allows autonomous agents (such as software programs) to implicitly learn the decision-making process of human drivers from experience.
Technical Details: The UCF invention comprises a Hybrid Predictive Network (HPN), a Value Function Network (VFN) and a safety prioritizer. Built on a symmetric encoder-decoder architecture, the HPN uses a series of observations to predict future scenarios. By combining the HPN's predictive capabilities with decision-making, the VFN estimates state-action value functions to improve navigation. A multi-step prediction chain uses the HPN to generate future hypotheses based on observation history. The safety prioritizer, integrated within the VFN, penalizes high-risk actions, masking them when selected, thus increasing safety.
Partnering Opportunity: The research team is seeking partners for licensing, research collaboration, or both.
Stage of Development: Prototype available.
Benefit
Enhances decision-making and safety of AV navigationCan prevent situations such as a vehicle not being allowed to merge on a highwayImproves traffic flow, reduces accidentsMarket Application
Autonomous vehicle and robotics manufacturersPublications
Prediction-aware and Reinforcement Learning based Altruistic Cooperative Driving, arXiv:2211.10585, submitted on 19 Nov 2022
Brochure