CoNav

CoNav是由华南理工大学等机构创建的协作导航数据集，专注于室内环境中的人机协作导航。该数据集包含超过25,000条多样化和真实的人形动画轨迹，这些轨迹是通过大型语言模型和生成模型结合环境上下文生成的。数据集的创建过程涉及环境对齐的活动推理、开放词汇动作动画和环境对齐的轨迹整合。CoNav数据集旨在解决现有导航方法忽视人类意图感知的问题，适用于家庭机器人、医疗辅助和老年人护理等多个应用领域，以提高人机协作的效率和效果。CoNav is a collaborative navigation dataset developed by institutions including South China University of Technology, focusing on human-robot collaborative navigation in indoor environments. This dataset contains over 25,000 diverse and realistic humanoid animation trajectories, which are generated by combining large language models (LLMs) and generative models with environmental context. The development of the CoNav dataset involves environment-aligned activity reasoning, open-vocabulary action animation, and environment-aligned trajectory integration. The CoNav dataset aims to address the issue that existing navigation methods overlook human intention perception, and is applicable to multiple application scenarios such as household robots, medical assistance, and elderly care to improve the efficiency and effectiveness of human-robot collaboration.

搜集汇总

背景与挑战

背景概述

CoNav是一个以人为中心的协作导航基准测试数据集，旨在使智能体能够通过观察人类活动来推断人类意图，并提前导航到目标位置。该数据集的核心特点是利用基于大语言模型（LLM）的人体动画生成框架，结合文本描述和环境上下文来创建现实且多样化的3D人类活动环境，以支持协作导航任务的研究和开发。