| Popis: |
Dynamic traffic control systems are important technical assets of the road infrastructure with regard to the efficiency and safety of freeway traffic on highly utilized roads. Based on distributed system architectures, they typically consist of numerous local sensors for measuring traffic flow and environmental conditions. Centralized and decentralized hardware and software components are responsible for data processing (including rule-based automated traffic control) and data communication. Human interaction in terms of manual control (as, for instance, in case of accident warnings) as well as continuous system monitoring is realized by operators in a traffic control center. Finally, from the viewpoint of the road users, the most visible components of such traffic control systems are the dynamic traffic signs used for displaying warnings (e.g., congestion, wet or icy road conditions, or accidents), speed limits, and possible restrictions on overtaking. Obviously, dynamic traffic control systems as described above are highly complex assets and thus difficult and expensive to maintain. Moreover, fault identification usually is an effortful manual process currently realized more or less systematically by experienced operators and maintenance engineers in the traffic control center and in the field. Model-based tools for automatic failure detection and diagnosis (i.e., identification of failure reasons) such as Bayesian networks provide the chance to significantly improve current maintenance practices including a possible shift from mostly corrective towards more condition-based and predictive maintenance. The present contribution discusses these potentials from a scientific as well as a practitioner's point of view including a critical review of current maintenance strategies and previous work on failure diagnostics for dynamic traffic control systems. |
