China
Africa
Explore chapters and articles related to this topic
Multi-Aerial-Robot Planning
Published in Yasmina Bestaoui Sebbane, Multi-UAV Planning and Task Allocation, 2020
A wide range of conflict resolutions have addressed the collision avoidance problem that can be split into three types: Prescribed: All aerial robots follow a set of protocols that tend to yield a discrete event controller, which when combined with the aerial robot’s continuous dynamics forms a hybrid system.Optimized: These methods attempt to find the best route for all the aerial robots to take to avoid each other while minimizing a cost function. Generally, they use a look ahead or time horizon so that the solution does not have to be recalculated often. The collision cone concept is a first order look ahead for detecting conflict. The collision cone (or velocity obstacle) is a set of velocities for one aerial robot that will cause it to collide with another, assuming each of their velocities are constant.Force field: These methods use continuous feedback mechanism to compute the control. The force field between two aerial robots is similar to the repulsion between two like charged particles. However, many possible alternatives are available for feedback schemes. These methods are generally reactive in that the control reacts to the current state of the system, rather than planning a trajectory ahead of time.
Special issue on the dynamics and behaviours of pedestrian groups
Published in Transportmetrica A: Transport Science, 2023
Tie-Qiao Tang, Alexandre Nicolas, Seungjae Lee, Winnie Daamen, Ziqi Song
In a study conducted by Han, Liu, and Li (2021), a guidance model is proposed to plan evacuation paths for pedestrians with limited vision. The proposed model considers the effects of three major objective factors – namely, the length of paths, the density of exits, and congestion – on pedestrians’ route choices. The distribution of pedestrians selecting different exits is defined as a feature of intermediate states during evacuations. Reinforcement learning (RL) is used to optimise the guidance model parameters. Evacuation efficiency is improved by optimising a series of intermediate states. A reciprocal velocity obstacle technology is used to build an evacuation model to verify the results of the study. Numerical simulations have shown that the guidance model can alleviate congestion, balance the utilisation of exits, and shorten evacuation time by guiding pedestrians’ route choices.
An evacuation guidance model for pedestrians with limited vision
Published in Transportmetrica A: Transport Science, 2023
Yanbin Han, Hong Liu, Liang Li
Many factors cause the limited vision of pedestrians, such as individual physiological factors, smoke, light, and the complexity of buildings. Limited vision seriously affects pedestrian judgment and the total evacuation time (TET). Evacuation guidance is an effective way to resolve the above problem. In this paper, we mainly focus on how to guide pedestrians with limited vision to shorten the TET. For this purpose, we propose a guidance model that contains three major objective factors affecting evacuation time: the length of routes, the congestion of routes, and the density of exits. Then, after analyzing the relationship between the mid-state and the evacuation time, we use reinforcement learning to optimize the guidance model. Finally, we use reciprocal velocity obstacle technology to build an evacuation model to verify our study. A series of simulation experiments illustrate that the guidance model can alleviate congestion, balance the utilization of exits, and shorten the evacuation time by guiding pedestrian evacuation.
Social groups in pedestrian crowds: review of their influence on the dynamics and their modelling
Published in Transportmetrica A: Transport Science, 2023
Alexandre Nicolas, Fadratul Hafinaz Hassan
In the case where collision avoidance is performed by a single individual facing a social group, the question had been addressed previously by (Bruneau, Olivier, and Pettre 2015). Using virtual reality experiments, they showed that participants tend to ‘go through’ large sparse groups, whereas they will ‘walk around’ small dense groups. For a circular group of radius 3 metres, the proportion of ‘go-through’ moves surges from around 20% to 100% when the interpersonal distance within the group increases from 1.1 m to 2.3 m. The extreme cases are well rationalised by a principle of minimum energy which compares the cost of a test go-through trajectory with that of a walk-around path. Unlike most of the aforementioned works, here, the trajectories were not computed with a force-based model, but with a velocity-obstacle model. In this branch of continuous models, for each agent, the space of possible velocities is explored in search of the best one in terms of an objective function, after excluding all velocities that are expected to lead to a collision (over a finite time horizon); the latter are called velocity-obstacles. (Ren et al. 2017) put forward a general-purpose method to introduce various types of groups within these models. It simply consists in further constraining the explored velocity space by barring velocities that drive an agent too far from her group neighbours at the next time step; a subtlety is that not all group members are regarded as neighbours, but only a subset of them defined by physical proximity and social relations. Finally, the explored velocity space may be overlaid with additional costs to favour specific group formations.