CrowdNav

该数据集名为CrowdNav，是一款旨在模拟机器人与人类共同参与的社交导航场景中复杂行人行为的模拟器。该模拟器包含了随机生成的人类位置和目标，同时为行人和机器人的大小及速度设定了特定参数。行人行为模型基于ORCA算法。该数据集的规模为20名行人和1台机器人，任务内容为社交导航。

This dataset, named CrowdNav, is a simulator designed to model complex pedestrian behaviors in social navigation scenarios involving collaborative participation of robots and humans. The simulator generates random initial positions and target waypoints for human pedestrians, and specifies dedicated parameters for the size and travel speed of both pedestrians and robots. The pedestrian behavior model is based on the ORCA algorithm. The dataset is configured for scenarios with 20 pedestrians and 1 robot, with the core task being social navigation.

搜集汇总

背景与挑战

背景概述

CrowdNav数据集用于评估SoNIC算法在社交导航中的性能，该算法通过结合自适应共形推理和约束强化学习，显著提高了机器人在人群中的安全性和社交礼仪表现，并在实际机器人部署中验证了其有效性。