MobileOcc: Unlocking Next-Gen Human-Aware 3D Perception for Mobile Robots

By Junseo Kim, Guido Dumont, Xinyu Gao, Gang Chen, Holger Caesar, Javier Alonso-Mora


Published on November 24, 2025| Vol. 1, Issue No. 1

Summary

MobileOcc is a novel semantic occupancy dataset specifically designed to enhance mobile robot navigation and perception in crowded, human-dense environments. Unlike existing datasets focused on autonomous driving, MobileOcc addresses the critical gap for mobile robots by providing rich 3D semantic occupancy data. It leverages an innovative annotation pipeline that combines static object occupancy with a mesh optimization framework to accurately model deformable human geometry using both 2D images and LiDAR data. The dataset establishes benchmarks for key tasks like occupancy prediction and pedestrian velocity prediction, offering various methods and baseline implementations, and has shown robust performance across different datasets.

Why It Matters

The introduction of MobileOcc marks a significant step forward for the practical deployment of mobile robots in human-centric environments, an area critically underserved by existing datasets primarily tailored for autonomous driving. For AI professionals, this dataset is more than just new data; it's an enabler for developing next-generation, truly human-aware AI systems in robotics. Better semantic occupancy and pedestrian velocity prediction are foundational for robots to navigate safely, efficiently, and acceptably among people, whether in factories, hospitals, or public spaces. This directly translates to enhanced safety, smoother human-robot interaction, and the ability for robots to perform complex tasks without causing disruption or harm. MobileOcc will accelerate research into more sophisticated perception models, allowing robots to anticipate human movements, understand intentions, and exhibit socially intelligent behaviors. Ultimately, by providing the crucial data needed to train robust AI, MobileOcc helps unlock the full potential of mobile robotics, moving them from controlled environments to dynamic, everyday settings, thereby expanding their utility and impact across numerous industries.

Advertisement