The concept of a system for searching for specific vehicles based on visual data, considered in articles [1,2,3,4], is being successively developed with further simulation stages. The primary goal is to achieve the highest probability of an accurate prediction of the target vehicle’s trajectory, which will ultimately allow for effective intervention by selected system users. Therefore, it is necessary to analyze diverse driver types as well as the accurate distribution of observation points. Currently, the results of research on a more elaborate driver model moving along a hypothetical road network have been presented. Distances from the starting point to the destination are measured along route segments, which can additionally be characterized by weights depending on road parameters (e.g., surface quality, capacity, division into toll roads, etc.). The effects of random deviations from optimal driver decisions (shortest routes) are taken into account when determining the positions of observation points. In the future, it is planned to optimize the quality of predictions, which will be presented in subsequent works.STUDY OF CHOSEN FACTORS INFLUENCING TRAJECTORIES OF SELECTED VEHICLES
Abstract
We extend the concept of a network for visual tracing definite vehicles, presented recently in [1,2,3,4], by successive improvements of available simulation components. The final aim is to make good predictions of future positions of the object under surveillance, which allows the interception of the target by the authoryties. It is crucial to include the whole spectrum of possible driver behaviors in order to decide reasonably about optimal positions for observation points. In the present article, the simulation is extended to a more flexible road net together with a more sophisticated driver model. Distances are measured along the road segments and may be weighted by road conditions such as state of the surface, traffic density and fees.The effects of random deviations from optimal decisions of drivers on the traffic densities at chosen camera positions are studied. Thus, the observation network may be subject to optimization with respect to the number of registered targets. Later on, the objective will be the quality of predictions, which will be considered in a forthcoming paper. (…)
The article contains 25,020 characters.
Source: Logistyka Magazine 6/2011
Vehicle search system, road network, simulation – PDF file for download









