Exploring Task Allocation and Safety Metrics for Autonomous Driving System Operations
/
Risk Institute researchers to present in the 34th European Safety and Reliability Conference (ESREL) in Cracow, Poland from June 23-27, 2024. These are some of the accepted papers:
"Task Allocation And Control Transitions In Autonomous Driving System Operations"
By: Camila Correa-Juillian, Marilia Ramos, Ali Mosleh, and Jiaqi Ma.
Abstract: Automated Driving Systems (ADS) are expected to play a significant role in the transportation environment in the coming decades, either deployed for passenger transport or as features in privately-owned vehicles. In both cases, humans will continue interacting with these systems as drivers, operators, and/or fellow road users. An element defining the current levels of driving automation is the task division and allocation between the human and the autonomous agent while operating under specific conditions. In this context, takeover and handover events, i.e., control transitions that can be triggered by exceeding the specified operational conditions, have become a focus of interest in multiple safety, reliability, and human factors research. This work discusses the high-level tasks human and autonomous agents perform in ADS operations. Three cases of interest are defined based on their relation to the ADS-equipped vehicle: a remote operator, a safety driver, and a consumer-level driver. This definition is based on which agent is responsible for high-level tasks, such as monitoring, planning, and executing the Dynamic Driving Tasks. A new taxonomy for control transitions and interventions is proposed for the three use cases. This taxonomy considers who initiated the control transition, who is in control after the transitions, the context that triggers the event, and whether it is a success or failure. Including successful or failed states in the taxonomy is relevant to address potential hazard scenarios and develop appropriate safety mechanisms to prevent or mitigate their risk.
“Exploring safety-related metrics to assess human-system interactions in heavy-duty automated vehicles”
By: Anna Cosmin-Spanoche, Camila Correa-Juillian, Xin Xia, Ali Mosleh, and Jiaqi Ma.
Abstract: Automated driving technologies are becoming increasingly common across various applications in the transportation industry. In recent years, there has been a growing interest in expanding these applications towards commercial heavy-duty operations, aiming to increase operational hours and reduce fatal collisions. Currently, multiple companies are involved in the development, testing, and small-scale deployment of heavy-duty automated vehicle (HD-AV) systems. With the emergence of new Automated Driving System (ADS) technology additional risks are introduced to commercial fleet operations. Currently, HD-AV fleet operations are planned as a joint effort of multiple human and machine agents, including an onboard safety driver and a fleet operations center. HD-AV operations can potentially cover a range of applications, including middle-mile, drayage, long-haul, etc. each with distinct operational profiles and safety requirements. In each of these, the interactions between agents contribute to the complexity of the operations and the design of safety requirements. Most notable among these are the interactions between the safety driver and the ADS, and these interactions must be modelled to construct an in-depth safety analysis. This work presents a discussion of current ADS and human-related safety metrics and suggests potential metrics that can build upon these to assess human-ADS interactions.
ESREL is the annual event in the area of reliability analysis, risk assessment, risk management and optimization of the safety performance of socio-technological systems.
