LHMP 2024: The THÖR-MAGNI challenge presentation by Janik Kaden (TU Chemnitz)

We are thrilled to announce the THÖR-MAGNI trajectory prediction benchmark and ICRA LHMP 2024 challenge.
The benchmark is set to test the generalization capabilities of motion prediction models in diverse indoor environments. To that end, we provide 5 scenarios with unique obstacle layouts and motion patterns and human tasks. Out of those, 4 will be used for training and validation, and the remaining one for testing. The benchmark uses the novel THÖR-MAGNI dataset (can be found at thor.oru.se), which is a large-scale multi-modal recording of human and robot motion in indoor environments.
Highlights of the benchmark:
* 5 scenarios with diverse obstacle maps and motion patterns
* Task-oriented navigation while interacting with and transporting various objects
* Social navigation between groups of people and a mobile robot
* Observation duration 3.2 seconds and prediction horizon 4.8 seconds
Available at:
github.com/sch...
github.com/tmr...
IEEE International Conference on Robotics and Automation (ICRA),
May 13, 2024
motionpredicti...