Showing papers on "Network traffic simulation published in 2021"

Synchronization and Feedback Loop Integration of a Non-real Time Microscopic Traffic Simulation with a Real-time Driving Simulator using Model-Based Prediction

Abstract: Driving simulators are a common tool to study human responses to driver assistance algorithms interacting with vehicle behavior. The accuracy of human responses, particularly to other vehicles, depends on the fidelity of representing the traffic surrounding the ego-vehicle. The traffic surrounding the ego- vehicle must also react to the ego-vehicle and be updated at a high enough rate such that the driver can perceive continuity. Microscopic traffic simulation tools are well developed to simulate traffic behavior replicating the real world over large networks. But the speed of microscopic simulations depends on the size of the network, traffic volume, simulation's computational hardware, and the traffic simulation software itself. The frequency of updating the traffic in many traffic simulators is not frequent enough for direct use in a driving simulator, thus causing discontinuous jumps in vehicle motion perceived by drivers. However, these traffic simulation results can be made continuous in perception by performing trajectory-level smoothing and time up-sampling as a real-time process occurring parallel to the driving simulator's rendering software. This ensures a realistic human perception of traffic behavior around the ego-vehicle. However, this integration creates the problem of synchronizing the traffic simulation dynamics with the ego-vehicle's dynamics within a driving simulator. This is challenging because traffic simulations are not typically programmed to be real-time, and thus the time offsets of the traffic simulator in relation to real-time can be time-varying. This paper describes a model-predictive feedback control method to synchronize a traffic simulator with a driving simulator in real-time by projecting the ego-vehicle into the future while adjusting time offsets via a feedback loop taking the traffic simulator's speed and communication delays into account. The update rate of the traffic is enhanced using a uniform acceleration observer that utilizes a localized simulation of the traffic network, and thus three dynamic systems are integrated at once: a large-scale network traffic simulation, a local realtime reduced -order model of the traffic, and the behavior of an ego-vehicle. This method of integrating traffic simulation with a driving simulator is demonstrated for a high-speed vehicle motion in a highway driving simulator, and the error analysis shows the success of this approach.

Systems and methods for manifolds learning of airline network data

Abstract: Methods and systems for using manifolds learning of airline network traffic to detect airline network traffic disruptions. In an embodiment, a computer processor (210) transmits a request for abnormal airline network traffic simulation data to a simulation device (206), receives the abnormal airline network traffic simulation data, and generates, utilizing a manifolds learning process, simulated abnormal airline network traffic manifolds representation data. In some embodiments, the process may next include transmitting a request for current airline network traffic data to a current network traffic computer (208), receiving the current airline network traffic data, and generating a current airline network traffic dataset. In some implementations, the current airline network traffic dataset is then mapped to the abnormal airline network traffic manifolds representation, a determination made that a threshold number of data points of the current airline network traffic dataset maps to the abnormal airline traffic manifolds representation, and then an alarm is transmitted to an operator device indicating that a disruption of airline network traffic is imminent.