Physica A: Statistical Mechanics and its Applications最新文献

Investigating pedestrian stepping is essential for pedestrian dynamics research, aiding in understanding pedestrian behavior and crowd modeling. However, how to calculate the basic step metrics is still controversial, and the differences between straight walking and turning steps are often overlooked in past studies. In this work, we proposed the trajectory-based measurement to more accurately calculate the step metrics and further analyze the differences between the straight walking and turning steps. The trajectory-based measurement takes the intrinsic trajectory of the pedestrian as the reference frame to guide a more universal measurement for stepping characteristics. By applying the proposed trajectory-based measurement to revisit the dataset of a single-file experiment, we identify differences between the straight walking step and the turning step from multiple perspectives. The results show that when density is low, straight walking steps exhibit larger step velocity and length, whereas turning steps display more unbalanced lateral motion. As density increases, both types of steps demonstrate greater forward motion imbalance, while pedestrians prefer to take the step on the outer side of the turn to propel their forward motion when taking turning steps. These findings deepen our understanding of pedestrian stepping behavior and provide valuable insights for future studies of pedestrian dynamics.

We investigate the effects of delayed interactions on the stationary distribution of the noisy voter model. We assume that the delayed interactions occur through the periodic polling mechanism and replace the original instantaneous two-agent interactions. In our analysis, we require that the polling period aligns with the delay in announcing poll outcomes. As expected, when the polling period is relatively short, the model with delayed interactions is almost equivalent to the original model. As the polling period increases, oscillatory behavior emerges, but the model with delayed interactions still converges to stationary distribution. The stationary distribution resembles a Beta-binomial distribution, with its shape parameters scaling with the polling period. The observed scaling behavior is non-monotonic. Namely, the shape parameters peak at some intermediate polling period.

Allowing connected and automated vehicles (CAVs) to drive in bus lanes can enhance the utilization efficiency of exclusive bus lanes. However, this could also lead to a certain degree of heightened bus operating delays. In a connected and automated environment, vehicle-to-vehicle (V2V) communication is essential for implementing dynamic management strategies for bus lanes. This study focuses on urban arterial roads as the research scenario. Based on the characteristics of the car-following and lane-changing behaviors of human-driven vehicles (HDVs) and CAVs, a simulation platform for heterogeneous traffic flow is established. Four strategies for bus lane control are proposed: exclusive bus lanes (EBLs), bus and CAV mixed lanes (BCMLs), dynamic bus lanes (DBLs), and dynamic priority bus lanes (DPBLs). Through simulation experiments, an analysis is conducted to compare traffic operation characteristics under different bus lane strategies, assess the advantages of these strategies, and determine the traffic density conditions for their implementation. The results indicate that in comparison to the EBL strategy, the other strategies enhance traffic flows but result in different levels of travel delays for buses. Compared to the DBL and BCML strategies, the DPBL strategy is applicable under a broader range of traffic density conditions. Additionally, sensitivity analysis revealed that the effects of the V2V communication distance and bus flows on the applicability of DPBL strategies are relatively minor. These results provide a theoretical basis and practical guidance for future bus lane management.

Connected and automated vehicle (CAV) provides a new promising solution for transportation system. Despite the promising future of CAV, fully deployment of CAV on current road systems is still challenging and the coexistence of CAV and human-driven vehicle (HDV) is inevitable. Furthermore, most studies for trajectory planning under mixed traffic ignore the stochasticity of human-driven vehicle (HDV), which is unrealistic and causes infeasible planned trajectory. In this study, we investigate merging control from a dedicated CAV lane into a conventional lane. The stochastic mixed traffic cooperative merging problem is formulated as a mixed integer quadratic constraint programming, which optimizes vehicle longitudinal trajectories and lane-changing maneuvers in a centralized way. Rolling horizon framework coupled with car-following and lane-changing execution algorithms is used to address the stochasticity of connected human-driven vehicle (CHV). Simulation results validate our proposed control strategy outperforms the rule-based control strategy from the perspective of traffic efficiency, lane-changing efficiency, fuel economy and driving comfort. The robustness of rolling horizon framework and sensitivity analysis are also conducted. Finally, the vehicle trajectory comparison intuitively shows the difference between 2 control strategies.

The self-organized criticality (SOC) exhibits excellent performance in describing the dynamic features of multi-body systems in nature. It is usually considered that conservative law is a necessary condition for the existence of SOC. However, mass or energy dissipation typically exists in the process from self-organization to a critical state in natural systems such as earthquakes and air pollution, namely, non-conservation. At present, few physical experiments have systematically explored the SOC behavior in a non-conservative system, compared to numerical simulations. This study is the first to design a non-conservative sandpile experiment using water droplets. Based on this experiment, the dynamic evolution process of avalanches in a non-conservative system has been analyzed. The results show that the avalanche size of water droplets follows a power-law distribution, as the water mass decays with time. This phenomenon obeys the rules of scale-free distribution, which is consistent with SOC behavior. Moreover, the waiting time of water droplets avalanche follows a stretched exponential distribution. These results suggest that SOC behavior still exists in non-conservative multi-body systems, which is affected by the decay coefficient. We then explore the mechanism by which the decay coefficient affects the transition of the system to a critical state in the non-conservative sandpile model (NBTW). We find that only when the decay coefficient is small do the inequality measures, Gini coefficient g and Kolkata index k, exhibit nearly universal values, similar to the mechanism of the system transitioning to a critical state in the traditional BTW model. When the decay coefficient is large, the NBTW model shows scaling-limited effects and the scaling exponent of avalanche size increases with the decay coefficient. Furthermore, we use the NBTW model to study how the decay coefficient influences the scaling exponent of avalanche size. We find that the NBTW model can effectively explain the variability of scaling exponents observed in different SOC systems and establish a quantitative relationship between the mass decay coefficient and the avalanche size scaling exponent. This study provides new insight into SOC behavior in non-conservative systems in nature.

Driving behavior modeling is extremely crucial for designing safe, intelligent, and personalized autonomous driving systems. In this paper, a modeling framework based on Markov Decision Processes (MDPs) is introduced that emulates drivers’ decision-making processes. The framework combines the Deep Maximum Entropy Inverse Reinforcement Learning (Deep MEIRL) and a reinforcement learning algorithm-proximal strategy optimization (PPO). A neural network structure is customized for Deep MEIRL, which uses the velocity of the ego vehicle, the pedestrian position, the velocity of surrounding vehicles, the lateral distance, the surrounding vehicles’ type, and the distance to the crosswalk to recover the nonlinear reward function. The dataset of drone-based video footage is collected in Xi’an (China) to train and validate the framework. The outcomes demonstrate that Deep MEIRL-PPO outperforms traditional modeling frameworks (Maximum Entropy Inverse Reinforcement Learning (MEIRL) - PPO) in modeling and predicting human driving behavior. Specifically, in predicting human driving behavior, Deep MEIRL-PPO outperforms MEIRL-PPO by 50.71% and 43.90% on the basis of the MAE and HD, respectively. Furthermore, it is discovered that Deep MEIRL-PPO accurately learns the behavior of human drivers avoiding potential conflicts when lines of sight are occluded. This research can contribute to aiding self-driving vehicles in learning human driving behavior and avoiding unforeseen risks.

The Sznajd model is one of sociophysics’s well-known opinion dynamics models. Based on social validation, it has found application in diverse social systems and remains an intriguing subject of study, particularly in scenarios where interacting agents deviate from prevailing norms. This paper investigates the generalized Sznajd model, featuring independent agents on a complete graph and a two-dimensional square lattice. Agents in the network act independently with a probability $p$, signifying a change in their opinion or state without external influence. This model defines a paired agent size $r$, influencing a neighboring agent size $n$ to adopt their opinion. This study incorporates analytical and numerical approaches, especially on the complete graph. Our results show that the macroscopic state of the system remains unaffected by the neighbor size $n$ but is contingent solely on the number of paired agents $r$. Additionally, the time required to reach a stationary state is inversely proportional to the number of neighboring agents $n$. For the two-dimensional square lattice, two critical points $p=p_c$ emerge based on the configuration of agents. The results indicate that the universality class of the model on the complete graph aligns with the mean-field Ising universality class. Furthermore, the universality class of the model on the two-dimensional square lattice, featuring two distinct configurations, is identical and falls within the two-dimensional Ising universality class.

As the demand for low-power electronics and IoT devices grows, ambient energy harvesting appears to be a promising alternative for powering such systems in the long run. However, optimizing power and efficiency concurrently in such systems is challenging, involving balancing a number of variables. This paper investigates the optimization of power and efficiency in ambient energy harvesting systems focusing on nonlinear oscillator electromechanical harvesters subjected to multiplicative time-correlated ambient noise. Through extensive numerical simulations, we reveal distinct relationships between power and efficiency, influenced by various parameters. We observe autonomous stochastic resonance phenomena, elucidating a linear power-efficiency trend for small noise correlation time under fixed noise variance but limiting simultaneous power and efficiency optimization beyond a threshold. Under fixed noise strength, there is a trade-off between power and efficiency. Additionally, damping strength, piezoelectric parameters, and capacitor charging time impact power and efficiency linearly. These insights enhance understanding of power efficiency dynamics in ambient energy harvesting, thereby offering practical recommendations for parameter selection to maximize both power output and efficiency in the next generation of electronics.

During the last years, cryptocurrencies have been increasingly becoming a relevant subject of academic researchers and investors. This paper adopts a novel framework that combines a multivariate Generalized AutoRegressive Conditional Heteroskedasticity (GARCH) and Copula modeling in a two-stage approach to analyze the cryptocurrency volatility dynamics. By combining the aforementioned techniques, on top of showing that price movements in one cryptocurrency can significantly influence others, the use of copulas highlight how these effects can vary across different parts of distributions and thus for different types of events with respect to their extreme nature. The interconnectedness complexity should be taken into consideration when managing risk in portfolio and constructing relevant models.