The development of vehicle assistance systems with active safety elements often requires repeated evaluation of both safety benefits and impacts of false positive activations within the design and optimization phases. The control parameters defining the operating point of an active safety system usually are designed to maximize effectiveness while minimizing false positives. Hence, the design of complex, active or "integral" safety systems generally requires methodologies for assessing both of these key characteristics. Moreover, the relative impacts of different kinds of false positive activations depend on the detailed system strategy: for example, superfluous warnings are far less hazardous in terms of controllability than false positive interventions.Deploying advanced driver assistance systems (ADAS) based purely on "engineering intuition" (without prior impact assessment) is neither risk-free nor cost-effective: Risks are associated with false positive interventions, for example. Moreover, long observation periods required to accumulate performance statistics result in a severe feedback lag to the development and optimization process. Thus, in order to design assistance systems that will most effectively reduce the number of accidents and their severity, there is an urgent need for reliable safety performance prediction during development, prior to deployment. In addition to automobile manufacturers and suppliers, public policy and opinion makers as well as regulatory agencies are key stakeholders in safety assessment. Assessment techniques likely to be accepted by all stakeholders should provide targeted, quantified safety performance prediction: In this context, a target corresponds to a specific traffic situation or accident scenario, for example, "pedestrian collisions involving a mid-block dash by the pedestrian". Quantification refers here to an objective metric: for example, the expected reduction in the MAIS2+ (maximum of the abbreviated injury scale) injuries to pedestrians. Further components of overall safety performance include estimated frequency of false positives (i.e., unnecessary triggering), classified according to severity of their side-effects.This paper describes virtual evaluation techniques such as those based on stochastic ("Monte-Carlo") simulations for analysis, assessment, and optimization of active and integral safety systems. In the stochastic simulation approach, all relevant natural and induced processes are modeled as a sequence of states subject to both deterministic and stochastic (random) influences and interactions. The methodology is illustrated for assessment of vehicle-based active pedestrian protection systems.
展开▼
