A kernel-based architecture for safe cooperative vehicular functions

Abstract

Future vehicular systems will be able to cooperate in order to perform many functions in a more effective and efficient way. However, achieving predictable and safe coordination of vehicles that autonomously cooperate in open and uncertain environments is a challenging task. Traditional solutions for achieving safety either impose restrictions on performance or require costly resources to deal with the worst case situations. In this paper, we describe a generic architectural pattern that addresses this problem. We consider that cooperative functions can be executed with multiple levels of service, and we rely on a safety kernel to manage the service level in run-time. A set of safety rules defined in design-time determine conditions under which the cooperative function can be performed safely in each level of service. The paper provides details of our implementation of this safety kernel, covering both hardware and software aspects. It also presents an example application of the proposed solutions in the development of a demonstrator using scaled vehicles.

VEDLIoT: Very Efficient Deep Learning in IoT

“…In brief, the approach consists in periodically submitting both the input and the output data to a robustness service, which holds a copy of the DL model and can verify the correctness of the output data. To support all these monitors and monitoring mechanisms, an architectural pattern comprising two separate parts is considered, based on the concept of architectural hybridization [16].…”

Cooperation with disagreement correction in the presence of communication failures

“…In this paper, we study a communication protocol that implements the KARYON's cooperative service level evaluator. In [7], we present the architecture that considers the interactions between the safety kernel, a local dynamic map and the cooperative service level evaluator. Unlike the earlier abstract presentation of the cooperative service level evaluator, this paper provides in detail, the design and analysis of the communication protocol.…”

“…Specifically, we address the following research question: How can cooperative systems be used to attain the highest performance without compromising safety in the presence of communication failures? We consider applications in which the individual vehicles estimate their ability to cooperate according to the sensory information quality and communicate their maximum supported cooperative level [5,7]. The vehicular system then decides on its cooperative service level according only to the received information.…”

How to Stop Disagreeing and Start Cooperatingin the Presence of Asymmetric Packet Loss

“…The application set of the proposed protocol includes distributed tasks that require: (i) repeated executions of uniform consensus and (ii) convergence within one communication round, but can (iii) tolerate disagreement (due to communication failures) for a minimal period. For example, we consider applications in which the individual vehicles estimate their ability to cooperate according to the sensory information quality and communicate their maximum supported cooperative operation mode according to [ 18 , 19 ]. The architectural component presented in [ 18 , 19 ] is called a safety kernel and it allows the vehicular system to use its sensory data to decide on its current level of service.…”

“…The algorithm presented here is the one that allowed the safety kernel of [ 18 , 19 ] to consider the data quality of all vehicles when taking this decision. In an earlier version of this work [ 20 ], we proposed the studied solution along with experiments for validating it.…”