Search

We argue that object detectors in the safety critical domain should prioritize detection of objects that are most likely to interfere with the actions of the autonomous actor. Especially, this applies to objects that can impact the actor’s safety and reliability. To quantify the impact of object (mis)detection on safety and reliability in the context of autonomous driving, we propose new object detection measures that reward the correct identification of objects that are most dangerous and most likely to affect driving decisions. To achieve this, we build an object criticality model to reward the detection of the objects based on proximity, orientation, and relative velocity with respect to the subject vehicle. Then, we apply our model on the recent autonomous driving dataset nuScenes, and we compare nine object detectors. Results show that, in several settings, object detectors that perform best according to the nuScenes ranking are not the preferable ones when the focus is shifted on safety and reliability.

Coding conventions are a means to improve the reliability of software systems, and they are especially useful to avoid the introduction of known bugs or security flaws. However, coding rules typically come in the form of text written in natural language, which makes them hard to manage and to enforce. Following the model-driven engineering principles, in this paper we propose an approach for the management and enforcement of coding conventions using structured models. We define the Coding Conventions Specification Language (CCSL), a language to define coding rules as structured specifications, from which checkers are derived automatically by code generation. To evaluate our approach, we run a thorough experiment on 8 real open-source projects and 77 coding rules for the Java language, comparing the violations identified by our checkers with those reported by the PMD static analysis tool. The obtained results are promising and confirm the feasibility of the approach. The experiment also revealed that textual coding rules rarely document all the necessary information to write a reliable checker.

Abstract Because of the substances they process and the conditions of operation, chemical plants are systems prone to the occurrence of undesirable and potentially dangerous events. Major accidents may occur when a triggering event produces a cascading accident that propagates to other units, a scenario known as domino effect. Assessing the probability of experiencing a domino effect and estimating the magnitude of its consequences is a complex task, as it depends on the nature of the substances being processed, the operating conditions, the failure proneness of equipment units, the execution of preventive maintenance activities, and of course the plant layout. In this work, we propose a stochastic modeling methodology to perform a probabilistic analysis of the likelihood of domino effects caused by propagating vapor cloud explosions. Our methodology combines mathematical models of the physical characteristics of the explosion, with stochastic state-based models representing the actual propagation among equipment units and the effect of maintenance activities. Altogether, the models allow predicting the likelihood of major events occurrence and the associated costs. A case study is analyzed, where various layouts of atmospheric gasoline tanks are assessed in terms of the predicted consequences of domino effects occurrence. The results of the analyses show that our approach can provide precious insights to support decision-making for safety and cost management.