loading page

Uncertainity and Noise Aware Decision Making for Autonomous Vehicles -A Bayesian Approach
  • +3
  • Rewat Sachdeva,
  • Raghav Gakhar,
  • Sharad Awasthi,
  • Kavinder Singh,
  • Ashutosh Pandey,
  • Anil Singh Parihar
Rewat Sachdeva
Delhi Technological University

Corresponding Author:[email protected]

Author Profile
Raghav Gakhar
Delhi Technological University
Sharad Awasthi
Delhi Technological University
Kavinder Singh
Department of Computer Science and Engineering at Delhi, Machine Learning Research Labora-tory, Technological University, Delhi Technological University
Ashutosh Pandey
Department of Computer Science and Engineering at Delhi, Machine Learning Research Laboratory, Technological University, Department of Computer Science and Engineering at Delhi, Machine Learning Research Labora-tory, Technological University, Delhi Technological University
Anil Singh Parihar
Department of Computer Science and Engineering at Delhi, Machine Learning Research Laboratory, Technological University, Department of Computer Science and Engineering at Delhi, Machine Learning Research Labora-tory, Technological University, Delhi Technological University

Abstract

In the evolving domain of autonomous vehicles, the importance of decision-making cannot be overstated. Deep Reinforcement Learning emerges as a pivotal tool in this landscape. However, traditional DRL algorithms grapple with inaccuracies in Q-value estimation, predominantly due to system noise and function approximation errors. Such inaccuracies, coupled with real-world unpredictabilities, often misdirect autonomous vehicles, jeopardizing safety. This work introduces a novel DRL algorithm tailored for uncertainty and noise-aware decisionmaking in autonomous vehicles. This novel approach harnesses Bayesian Neural Networks (BNN) and skew-geometric Jensen-Shannon divergence, rectifying the aforementioned limitations and also improving exploration. Evaluated in the OpenAI gymnasium environment, the algorithm has clear advantages over contemporary methods in terms of cumulative rewards and convergence speed.
30 Dec 2023Submitted to TechRxiv
08 Jan 2024Published in TechRxiv