While autonomous vehicles (AVs) continue to advance and reshape modern transportation, they remain unable to navigate all traffic conditions without human input, highlighting the need for remote human intervention in edge-case scenarios. Two major teleoperation paradigms have emerged to address this need: tele-driving and tele-assistance. In tele-driving, remote operators (ROs) continuously control the AV through direct access to its actuators. In tele-assistance, ROs provide high-level instructions through a specialized interface, with low-level maneuvers delegated to the AV. We conducted a quantitative comparison of these paradigms, examining four edge-case scenarios: one uses a steering wheel and pedals, the other employs discrete high-level commands through a Wizard-of-Oz methodology. We measured mental workload, situation awareness, time completion, and overall user experience (UX). Results indicate that the tele-assistance interface reduced operators’ mental workload and improved situation awareness, suggesting the need for further development of tele-assistance interfaces.

Autonomous vehicles (AVs) are rapidly evolving as a novel way of transportation. Nevertheless, there is a consensus that AVs cannot address all traffic scenarios independently. Consequently, there arises a need for remote human intervention. To pave the way for large-scale deployment of AVs onto public roadways, innovative models of remote operation must evolve. Such a paradigm is Tele-assistance, which posits that the low-level control of AVs should be delegated through high-level commands. Our work explores how such a command language should be constructed as a first step in designing a Tele-assistance user interface. Through interviews with 17 experienced teleoperators, we elicit a set of discrete commands that a remote operator can use to resolve various road scenarios. Subsequently, we create a scenario-command mapping and a thematic classification of the defined commands. Finally, we present an initial Tele-assistance interface design based on these commands.

Autonomous vehicle (AV) technologies are rapidly evolving with the vision of having self-driving cars moving safely with no human input. However, it is clear that at least in the near and foreseeable future, AVs will not be able to resolve all road incidents and that in some situations remote human assistance will be required. However, remote driving is not trivial and introduces many challenges stemming mostly from the physical disconnect of the remote operator. In order to highlight these challenges and understand how to better design AV teleoperation interfaces, we conducted several observations of AV teleoperation sessions as well as in-depth interviews with 14 experts. Based on these interviews, we provide an investigation and analysis of the major AV teleoperation challenges. We follow this by providing design suggestions for the development of future teleoperation interfaces for assistance and driving of AVs.