Skip to Main content Skip to Navigation

Autonomic Framework For Safety Management In The Autonomous Vehicle

Abstract : The development and deployment of Autonomous Vehicles (AV) is a very challenging endeavour from a safety perspective. These vehicles are safety-critical systems must navigate through multiple complex situations preventing any potential harm and without disturbing traffic flow to be accepted by society. Safe driving under full computer control also requires to interact and operate around different entities within complex road networks and appropriately address their various behaviours. While much progress has been achieved within the past years, work has centred on providing vehicles with the ability to navigate autonomously. Safety has emerged as the major challenge, not only to manage malfunctions or external disturbances but also on the vehicle behavioural part to address edge-cases. This thesis addresses the research question of how safe autonomy is formulated and managed in the literature. We review safety mitigation mechanisms at run-time employing adaptive behaviours. We identify that AV systems require a handful combination of observability, traceability, reconfigurability and flexibility. Based on these non-functional properties, we propose a framework that incorporates the notion of self-safety into existing AVs a manageable and scalable manner. The framework defines our methodology to represent the safety argumentation as constraints and our reference architecture that involve two layers that operate self-adaptation mechanisms to ensure safety. The first layer is closer to the autonomous vehicle and consists of a collection of dependable processes. They specify requirements and are coupled to control loops to manage the assurance of safety closely. The second layer reconfigures the workflow of the previous layer according the association between the constraints and the requirements. The control loops operate with respect upon the context according to the context-dependence restrictions as well as the state of the AV functions. We also detail the constituent parts and application of the framework, namely, with knowledge representation, abstractions, templates and the mechanisms that connect the control-loops as composable and agnostic microservices. This novel formulation is applied to a use case relating pedestrians, thus describing how the proposed safety approach can be implemented and tested. Results analysis and discussion on the perspectives are included.
Complete list of metadata

Cited literature [164 references]  Display  Hide  Download
Contributor : Matthieu CARRE Connect in order to contact the contributor
Submitted on : Saturday, January 25, 2020 - 7:00:17 PM
Last modification on : Tuesday, February 15, 2022 - 3:41:27 AM


Files produced by the author(s)


  • HAL Id : tel-02455266, version 1



Matthieu Carre. Autonomic Framework For Safety Management In The Autonomous Vehicle. Automatic. Université de Pau et des Pays de l'Adour, 2019. English. ⟨tel-02455266⟩



Record views


Files downloads