Measuring air traffic complexity based on small samples

Air traffic complexity is an objective metric for evaluating the operational condition of the airspace. It has several applications, such as airspace design and traffic flow management.Therefore, identifying a reliable method to accurately measure traffic complexity is important. Considering that many factors correlate with traffic complexity in complicated nonlinear ways,researchers have proposed several complexity evaluation methods based on machine learning models which were trained with large samples. However, the high cost of sample collection usually results in limited training set. In this paper, an ensemble learning model is proposed for measuring air traffic complexity within a sector based on small samples. To exploit the classification information within each factor, multiple diverse factor subsets（FSSs） are generated under guidance from factor noise and independence analysis. Then, a base complexity evaluator is built corresponding to each FSS. The final complexity evaluation result is obtained by integrating all results from the base evaluators. Experimental studies using real-world air traffic operation data demonstrate the advantages of our model for small-sample-based traffic complexity evaluation over other stateof-the-art methods.

Survivability evaluation for networks carrying complex traffic flows

The capability of a system to fulfill its mission promptly in the presence of attacks,failures,or accidents is one of the qualitative definitions of survivability.In this paper,we propose a model for survivability quantification,which is acceptable for networks carrying complex traffic flows.Complex network traffic is considered as general multi-rate,heterogeneous traffic,where the individual bandwidth demands may aggregate in complex,nonlinear ways.Blocking probability is the chosen measure for survivability analysis.We study an arbitrary topology and some other known topologies for the network.Independent and dependent failure scenarios as well as deterministic and random traffic models are investigated.Finally,we provide survivability evaluation results for different network configurations.The results show that by using about 50%of the link capacity in networks with a relatively high number of links,the blocking probability remains near zero in the case of a limited number of failures.

Genetic Algorithm‑Based SOTIF Scenario Construction for Complex Traffic Flow

The Safety of The Intended Functionality(SOTIF)challenge represents the triggering condition by elements of a specific scenario and exposes the function limitation of an autonomous vehicle(AV),which leads to hazards.As for operationcontent-related features,the scenario is similar to AVs’SOTIF research and development.Therefore,scenario generation is a significant topic for SOTIF verification and validation procedure,especially in the simulation testing of AVs.Thus,in this paper,a well-designed scenario architecture is first defined,with comprehensive scenario elements,to present SOTIF trigger conditions.Then,considering complex traffic disturbance as trigger conditions,a novel SOTIF scenario generation method is developed.An indicator,also known as Scenario Potential Risk,is defined as the combination of the safety control intensity and the prior collision probability.This indicator helps identify critical scenarios in the proposed method.In addition,the corresponding vehicle motion models are established for general straight roads,curved roads,and safety assessment areas.As for the traffic participants’motion model,it is designed to construct the key dynamic events.To efficiently search for critical scenarios with the trigger of complex traffic flow,this scenario is encoded as genes and it is regenerated through selection,mutation,and crossover iteration processes,known as the Genetic Algorithm(GA).Experimental results show that the GA-based method could efficiently construct diverse and critical traffic scenarios,contributing to the construction of the SOTIF scenario library.

Autonomous driving in the uncertain traffic--a deep reinforcement learning approach

representation capability of deep learning(DL) and the optimal decision making and control capability of reinforcement learning(RL), is a good approach to address this problem. Traffic environment is built up by combining intelligent driver model(IDM) and lane-change model as behavioral model for vehicles. To increase the stochastic of the established traffic environment, tricks such as defining a speed distribution with cutoff for traffic cars and using various politeness factors to represent distinguished lane-change style, are taken. For training an artificial agent to achieve successful strategies that lead to the greatest long-term rewards and sophisticated maneuver, deep deterministic policy gradient(DDPG) algorithm is deployed for learning. Reward function is designed to get a trade-off between the vehicle speed, stability and driving safety. Results show that the proposed approach can achieve good autonomous maneuvering in a scenario of complex traffic behavior through interaction with the environment.