Won first place in BARN Challenge 2024 hardware competition
May 2024 (202 Words, 2 Minutes)
BARN Challenge 2024 Hardware Competition Results
After securing first place in the simulation competition, our LiCS-KI team advanced to the physical finals of the BARN Challenge 2024 at ICRA in Yokohama. We’re proud to announce that we won first place in the hardware competition, successfully completing 6 out of 9 trials in real-world obstacle courses.
Hardware Competition Results
The physical competition involved navigating a real Clearpath Jackal robot through three different obstacle courses made of cardboard boxes. Each team had 30 minutes to complete five timed trials per course, with the top three trials counted. Success was measured by reaching the goal without any collisions.
| Rangking | Team | Success / Total Trials | Institution |
|---|---|---|---|
| 1 | LiCS-KI | 6/9 | Korea Advanced Institute of Science and Technology |
| 2 | MLDA_EEE | 5/9 (78.8s) | Nanyang Technological University, Singapore |
| 3 | AIMS | 5/9 (108.8s) | The Hong Kong Polytechnic University |
Source: BARN Challenge 2024 Official Results (Archive)
About the Physical Competition
The hardware competition tested the real-world capabilities of navigation systems that performed well in simulation. Key challenges included:
- Sim-to-Real Gap: Dealing with real-world factors like wheel friction, motor dynamics, and sensor noise
- Time Pressure: 30 minutes per obstacle course with multiple trial attempts
- Physical Constraints: Navigating through tightly constrained spaces with cardboard obstacles
- Onboard Computation: Limited to Intel i3 CPU with 16GB RAM (no GPU available)
Our Achievement
Winning both the simulation and hardware competitions demonstrates the robustness of our LiCS (Learned-imitation on Cluttered Space) approach. Our behavior cloning method with Transformer neural networks proved effective not only in simulation but also in real-world scenarios with all their inherent uncertainties.
The victory showcases the practical applicability of our navigation system for real-world robotics applications, from search and rescue operations to autonomous delivery in constrained environments.
Competition Format
- Three obstacle courses with different layouts and difficulty levels
- Five timed trials per course (30 minutes total per course)
- Top three trials counted for final scoring
- Success criteria: Reach goal without collision
- Tiebreaker: Fastest average traversal time for teams with equal success rates
This dual victory in both simulation and hardware validates our approach’s effectiveness across the complete development pipeline from virtual testing to real-world deployment.
Related Publications
-
LiCS: Navigation using Learned-imitation on Cluttered Space
Joshua Julian Damanik, Jae-Won Jung, Chala Adane Deresa, Han-Lim Choi -
Autonomous Ground Navigation in Highly Constrained Spaces: Lessons Learned From the Third BARN Challenge at ICRA 2024 [Competitions]
Xuesu Xiao , Zifan Xu, Aniket Datar, Garrett Warnell, Peter Stone, Joshua Julian Damanik, Jaewon Jung, Chala Adane Deresa, Than Duc Huy, Chen Jinyu, Chen Yichen, Joshua Adrian Cahyono, Jingda Wu, Longfei Mo, Mingyang Lv, Bowen Lan, Qingyang Meng, Weizhi Tao, and Li Cheng