THÖR-MAGNI is a public dataset of human motion trajectories, recorded in a controlled indoor experiment.
We present THÖR-MAGNI: a dataset of motion trajectories aligned with eye tracking data with diverse and accurate social human motion data in a shared indoor environment. We provide five distinct scenarios, that enable users to study various human activities and human motion in the presence of navigating robots and during interactions with them.
The dataset was recorded in a spacious laboratory room of 8.4x18.8 m. The laboratory room, where the motion capture system is installed, is mostly empty to allow for maneuvering of large groups, but also includes several constrained areas where obstacle avoidance and the choice of homotopy class is necessary. Goal positions are placed to force navigation along the room and generate frequent interactions in its center, while the placement of obstacles prevents walking between goals on a straight line.
To track the motion of the agents we used the Qualisys Oqus 7+ motion capture system with 10 infrared cameras, mounted on the perimeter of the room. For people tracking, the reflective markers have been arranged in distinctive 3D patterns on the bicycle helmets. There are 10 helmets in this dataset, marked from 1 to 10.
For recording gaze directions, we used two Tobii Pro Glasses (Model 2 and 3) and a Pupil Core Invisble mobile tracking headset. The gaze sampling frequency of Tobii Pro Glasses and the Pupil Core Invisible is 50 Hz. All three headsets have a scene camera which records the video at 25 fps for the Tobii- and 30 Hz for Pupil Glasses. A gaze overlaid version of these videos is included in this dataset. Additionally we include the 2D- (For all devices) and 3D (For the Tobii Devices) eye tracking data aligned with the motion capture systems’ data.
Our environment features several robotic arms as static obstacles and a mobile robot called DARKO. DARKO has a base with omnidirectional navigation capabilities. It is equipped with a robotic arm on top. The robot base is RB-Kairos+ and the arm is the Collaborative Robot Panda from Franka Emika. The robot base dimensions are 760×665×690 mm. The maximum reach height of the robot arm is 855 mm. The robot has one Ouster OS0-128 LiDAR, two Azure Kinect RGB-D cameras (one used in these recordings), two Basler fish-eye RGB cameras, and two Sick MicroScan 2D safety LiDARs. The Azure Kinect camera has a 75-degree horizontal field of view and a tracking range of up to 5 m. Finally, the robot is equipped with an “Anthropomorphic robot Mock Driver” (ARMoD). In scenarios 3,4 and 5, DARKO is a moving agent and plays an important role to study human-robot interaction settings.
In order to collect motion data applicable to a variety of research areas, we designed several scenarios that promoted social interactions between individuals, groups of people, and robots. Within a segment of the recordings, participants engaged in activities (roles) specifically designed to emulate industrial tasks. These activities include both group and individual movements, as well as the transportation of various objects, including items such as boxes, buckets, and larger objects such as a poster stand. The goal is to motivate participants to move heterogeneously and to create meaningful, rich, and natural interactions given by the different human roles.
THÖR-MAGNI dataset is a novel data collection of accurate human and robot navigation and interaction in diverse indoor contexts, building on the previous THÖR dataset protocol. We provide position and head orientation motion capture data, eye gaze tracking, 3D LiDAR scans and RGB data onboard a mobile robot. In total, THÖR-MAGNI captures 3.5 hours of motion of 40 participants on 5 recording days. This data collection is designed around systematic variation of factors in the environment to allow building cue-conditioned models of human motion and verifying hypotheses on factor impact.
Additionally, we provide a set of data visualization tools, including a dashboard, and introduce a specialized Python package thor-magni-tools. This package is designed to facilitate the filtering and preprocessing of raw trajectory data, enhancing the accessibility and usability of the THÖR-MAGNI dataset.
@article{schreiter2022magni, title={The magni human motion dataset: Accurate, complex, multi-modal, natural, semantically-rich and contextualized}, author={Schreiter, Tim and de Almeida, Tiago Rodrigues and Zhu, Yufei and Maestro, Eduardo Gutierrez and Morillo-Mendez, Lucas and Rudenko, Andrey and Kucner, Tomasz P and Mozos, Oscar Martinez and Magnusson, Martin and Palmieri, Luigi and others}, journal={arXiv preprint arXiv:2208.14925}, year={2022} }
This work was partially supported by Darko Project, an international research project on dynamic agile production robots that learn and optimise knowledge and operations.