Exiting the Simulation: The Road to Robust and Resilient Autonomous Vehicles at Scale

Richard Chakra

DOI: https://doi.org/10.48550/arXiv.2210.10876

2022-10-20

Abstract:In the past two decades, autonomous driving has been catalyzed into reality by the growing capabilities of machine learning. This paradigm shift possesses significant potential to transform the future of mobility and reshape our society as a whole. With the recent advances in perception, planning, and control capabilities, autonomous driving technologies are being rolled out for public trials, yet we remain far from being able to rigorously ensure the resilient operations of these systems across the long-tailed nature of the driving environment. Given the limitations of real-world testing, autonomous vehicle simulation stands as the critical component in exploring the edge of autonomous driving capabilities, developing the robust behaviors required for successful real-world operation, and enabling the extraction of hidden risks from these complex systems prior to deployment. This paper presents the current state-of-the-art simulation frameworks and methodologies used in the development of autonomous driving systems, with a focus on outlining how simulation is used to build the resiliency required for real-world operation and the methods developed to bridge the gap between simulation and reality. A synthesis of the key challenges surrounding autonomous driving simulation is presented, specifically highlighting the opportunities to further advance the ability to continuously learn in simulation and effectively transfer the learning into the real-world - enabling autonomous vehicles to exit the guardrails of simulation and deliver robust and resilient operations at scale.

Robotics,Artificial Intelligence

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to ensure the robust and resilient operation of autonomous vehicles (AVs) before large - scale practical deployment. Although in recent years, autonomous driving technology has made remarkable progress in perception, planning and control capabilities and has begun public road trials, it is still a great challenge to strictly ensure that these systems can operate robustly in the long - tailed distribution of the driving environment. Due to the limitations of real - world testing, autonomous vehicle simulation has become a key component for exploring the limits of autonomous driving capabilities, developing the required robust behaviors, and extracting hidden risks from complex systems before deployment. This article aims to introduce the current state - of - the - art autonomous driving system simulation frameworks and methods, with a focus on how simulation can be used to build the resilience required for real - world operations and how the developed methods can bridge the gap between simulation and reality. In addition, the article also synthesizes the key challenges surrounding autonomous driving simulation, especially emphasizing the opportunity to further enhance the ability to continuously learn in simulation and effectively transfer this learning outcome to the real world, enabling autonomous vehicles to break free from the limitations of simulation and achieve large - scale robust and resilient operations.