Autonomous road vehicles hold the promise of making road transportation potentially safer, more fluid and more sustainable. Clearly, the above-mentioned goals imply stringent reliability and maintainability targets on the roadside detectors.

The present paper reports on two complementary studies on the latter: (1) focusing on true detection performance, characterized by true detection probability, i.e. the ability for a roadside detector to detect soon enough a fixed obstacle lying on the road, thus avoiding potentially hazardous emergency brakes; (2) the other, on so-called false positives, i.e. the possibility for a roadside detector to signal the presence of an object when there is none.

In both cases, quantitative performance targets lead to reliability allocations to the detectors and to constraints exported to maintainer and operator.

For the true detection capability, or `true positive rate', using standard probabilistic assumptions, a closed-form constraint is obtained which relates the minimum acceptable true-positive target, the maximum acceptable detector failure rate and the average mileage driven annually. Reliability allocations can then be made to the detectors under various maintenance policy assumptions.

For the false positives, i.e. spurious detection of an obstacle ahead of an autonomous vehicle, there is (i) a safety concern: the vehicle may not have time to brake smoothly; and (ii) an availability concern: a slowdown of one or more vehicles may result in a "traffic jam''. To keep those risks acceptable, constraints are formulated in terms of kinematic parameters and detectors density but also average traffic supervision operator reaction time.