Forerunner UAV refers to an unmanned aerial vehicle equipped with a downward-looking camera flying in front of the advancing emergency ground vehicles (EGV) to notify the driver about the hidden dangers (e.g., other vehicles). A feasibility demonstration in an urban environment having a multicopter as the forerunner UAV and two cars as the emergency and dangerous ground vehicles was done in ZalaZONE Proving Ground, Hungary. After the description of system hardware and software components, test scenarios, object detection and tracking, the main contribution of the paper is the development and evaluation of encounter risk decision methods. First, the basic collision risk evaluation applied in the demonstration is summarized, then the detailed development of an improved method is presented. It starts with the comparison of different velocity and acceleration estimation methods. Then, vehicle motion prediction is conducted, considering estimated data and its uncertainty. The prediction time horizon is determined based on actual EGV speed and so braking time. If the predicted trajectories intersect, then the EGV driver is notified about the danger. Some special relations between EGV and the other vehicle are also handled. Tuning and comparison of basic and improved methods is done based on real data from the demonstration. The improved method can notify the driver longer, identify special relations between the vehicles and it is adaptive considering actual EGV speed and EGV braking characteristics; therefore, it is selected for future application.

Contextual calibration for object detection is a technique where a pretrained network collects attractive false positives during a calibration phase and use this calibration data for further training. This paper investigates the applicability of this method to a vision based onboard sense and avoid system, which requires intruder aircraft detection in camera images. Various landscape and sky backgrounds were generated by Unreal4 3D engine for calibration tests. Contextual calibration is a promising candidate for handling extreme situations which are not covered well in the training data.