Search

To avoid crowd evacuation simulations depending on 2D environments and real data, we propose a framework for crowd evacuation modeling and simulation by applying deep reinforcement learning (DRL) and 3D physical environments (3DPEs). In 3DPEs, we construct simulation scenarios from the aspects of geometry, semantics and physics, which include the environment, the agents and their interactions, and provide training samples for DRL. In DRL, we design a double branch feature extraction combined actor and critic network as the DRL policy and value function and use a clipped surrogate objective with polynomial decay to update the policy. With a unified configuration, we conduct evacuation simulations. In scenarios with one exit, we reproduce and verify the bottleneck effect of congested crowds and explore the impact of exit width and agent characteristics (number, mass and height) on evacuation. In scenarios with two exits and a uniform (nonuniform) distribution of agents, we explore the impact of exit characteristics (width and relative position) and agent characteristics (height, initial location and distribution) on agent exit selection and evacuation. Overall, interactive 3DPEs and unified DRL enable agents to adapt to different evacuation scenarios to simulate crowd evacuation and explore the laws of crowd evacuation. [ABSTRACT FROM AUTHOR]

To overcome the difficulties of quantitatively optimizing and evaluating evacuation guidance systems, we proposed a perception model based on virtual reality (VR) experiments and the social force model (SFM). We used VR and eye tracking devices to carry out experiments. The VR experiment data was mainly used for three purposes: to determine the parameter values of the perception model, to optimize the evacuation guidance system by quantitative analysis, and to validate the perception model. Additionally, we compared the VR experimental and model simulation results before and after the optimization to quantitatively assess the improvement in the optimized evacuation guidance system. The results showed that our model can effectively simulate the perception behaviors of evacuees on the evacuation guidance system and it can quantitatively evaluate different evacuation guidance system schemes. The model simulations showed that the optimized evacuation guidance system improved the evacuation efficiency, with the average escape time and distance of the two starting positions reduced by 37% and 28%, respectively. [ABSTRACT FROM AUTHOR]

At present, a common drawback of crowd simulation models is that they are mainly simulated in (abstract) 2D environments, which limits the simulation of crowd behaviors observed in real 3D environments. Therefore, we propose a deep reinforcement learning-based model with human-like perceptron and policy for crowd evacuation in 3D environments (HDRLM3D). In HDRLM3D, we propose a vision-like ray perceptron (VLRP) and combine it with a redesigned global (or local) perceptron (GOLP) to form a human-like perception model. We propose a double-branch feature extraction and decision network (DBFED-Net) as the policy, which can extract features and make behavioral decisions. Moreover, we validate our method's ability to reproduce typical phenomena and behaviors through experiments in two different scenarios. In scenario I, we reproduce the bottleneck effect of crowds and verify the effectiveness and advantages of HDRLM3D by comparing it with real crowd experiments and classical methods in terms of density maps, fundamental diagrams, and evacuation times. In scenario II, we reproduce agents' navigation and obstacle avoidance behaviors and demonstrate the advantages of HDRLM3D for crowd simulation in unknown 3D environments by comparing it with other deep reinforcement learning-based models in terms of trajectories and numbers of collisions. [ABSTRACT FROM AUTHOR]

In order to study the crowd evacuation process in complex subway stations, this paper proposes a social group perception evacuation model based on the social group evacuation model and the perception evacuation model, combined with the spatial index algorithm and the social group random algorithm, and takes a real subway station as an example to simulate the crowd evacuation process. Through the simulation results, the evacuation bottlenecks in the subway station are analyzed and the influence of social groups on evacuation efficiency is also verified. The results show that subway gates, fenced aisles and exits are important bottlenecks for crowd evacuation in subway stations, and the social groups will reduce the evacuation efficiency. This model can more truly restore the interaction behavior between evacuees and the environment in the evacuation process, and is suitable for the real complex subway stations. • Bottlenecks that affecting evacuation efficiency in subway stations mainly include subway gates, fenced aisles and exits. • Social groups have an important effect on evacuation efficiency. • The variation trends of panic index of individuals and social groups under different desired velocities are basically the same. [ABSTRACT FROM AUTHOR]

A new super queue model with yielding behavior is proposed to model the orderly evacuation process of young pupils in a classroom building. A super queue is composed of a super leader and several simple queues. Based on this, a leader-follower model is proposed to model the orderly behavior between pupils in a super queue, and an exclusive channel model is proposed to model the yielding behavior between super queues. Taking the evacuation of a classroom building as an example, the super queues and yielding behavior are reproduced and are very similar to the actual behavior (both qualitatively and quantitatively). Based on the model, scenarios are simulated to analyze the impacts of several factors on evacuation efficiency. The results show the following: (1) The stop-and-go effect greatly affects the overall evacuation efficiency. Therefore, it is necessary for students to maintain sufficient concentration to shorten their reaction time during the stopping and going process, thereby improving the evacuation efficiency. (2) Different yielding strategies have little impact on the overall evacuation efficiency. Therefore, in an actual evacuation case, flexible yielding strategies can be adopted in order to consider the needs of each class. (3) By optimizing evacuees' behaviors, such as improving their reaction times, the efficiency of an evacuation with queuing and yielding can become close to that of a disordered evacuation. At the same time, due to its higher safety and easier management process, a queuing and yielding evacuation can be a safe and relatively efficient method for real evacuations. [ABSTRACT FROM AUTHOR]