LocoVR: Multiuser Indoor Locomotion Dataset in Virtual Reality
Kojiro Takeyama, Yimeng Liu, Misha Sra

TL;DR
LocoVR is a comprehensive VR-based dataset of over 7000 human trajectories in indoor home environments, capturing social navigation behaviors crucial for developing socially-aware AI agents.
Contribution
This paper introduces LocoVR, a large-scale dataset with nuanced social navigation behaviors in indoor settings, filling gaps in existing human motion datasets.
Findings
LocoVR improves model performance in indoor navigation tasks.
The dataset captures socially-aware movement behaviors.
Models trained on LocoVR better predict human trajectories in home environments.
Abstract
Understanding human locomotion is crucial for AI agents such as robots, particularly in complex indoor home environments. Modeling human trajectories in these spaces requires insight into how individuals maneuver around physical obstacles and manage social navigation dynamics. These dynamics include subtle behaviors influenced by proxemics - the social use of space, such as stepping aside to allow others to pass or choosing longer routes to avoid collisions. Previous research has developed datasets of human motion in indoor scenes, but these are often limited in scale and lack the nuanced social navigation dynamics common in home environments. To address this, we present LocoVR, a dataset of 7000+ two-person trajectories captured in virtual reality from over 130 different indoor home environments. LocoVR provides accurate trajectory data and precise spatial information, along with rich…
Peer Reviews
Decision·ICLR 2025 Poster
+ The dataset is based on real human subject studies, and the use of VR environments facilitates data collection efforts. + This open-source dataset could benefit the research community focused on social navigation.
- Novelty and contributions are key concerns. Although the human subject studies require significant time and the dataset a reasonable sample size, the dataset has not been demonstrated, for example, to train state-of-the-art neural network models. The demonstrated methods are relatively simple. The dataset is also limited to two-person navigation scenarios. - Given that the objective of the dataset is to estimate user trajectories and goal positions, without addressing the estimation of human
1. The extensive collection of two-person human trajectory (with motion capture) data in indoor scenes, is a valuable resource for the community. 2. Utilizing virtual reality for data collection is a promising approach, allowing more diverse 3D scenes when resources are limited. 3. The data and code will be released.
1. While using VR to collect human trajectory data is helpful, this paper would benefit from a discussion in the related works section about VR and human motion. For instance, referencing works like "QuestEnvSim: Environment-aware Simulated Motion Tracking from Sparse Data" in SIGGRAPH 2023 which uses VR for motion tracking and "Strategy and Skill Learning for Physics-based Table Tennis Animation" in SIGGRAPH 2024 which involves interaction between human and humanoid agents. 2. I notice authors
1) LocoVR’s large, indoor locomotion dataset records two-person interactions across 130 diverse indoor environments 2) Introduction of motion proxemics in a large-scale dataset which is useful for downstream tasks such as studying human-human interactions and potentially human-robot interactions 3) Rigorously quantitatively evaluated against strong baselines; baseline configurations and settings are fair and well-documented 4) Data of two-person trajectories is useful to study motion proxemics
1) Potential Overfitting to VR-Specific Biases: I am curious what the authors have done to further minimize the gap between real-world scenes vs VR scenes. Are there any obstacle perception features e.g., vibrational feedback when participants bump into objects in the scene? 2) *Qualitative results: ‘as the trajectory progresses, the probability distribution of the goal area narrows down near the true goal object’* I think that it is reasonable to assume that humans narrow the probability dist
Code & Models
Videos
Taxonomy
TopicsVideo Surveillance and Tracking Methods · Indoor and Outdoor Localization Technologies · IoT-based Smart Home Systems
