MLDA Robo Team for BARN Challenge ICRA 2024

During my final year in Machine Learning and Data Analytics Club (MLDA@EEE), I led the Robotics Team initiative, gathering a group of EEE friends who are interested in Robotics participate in the BARN Challenge (Benchmark Autonomous Robot Navigation) at ICRA 2024.

Compared to Garage@EEE, MLDA@EEE is a software-focused club. Hence, our robotics projects require commercially available robot hardware or simulated environment.

BARN Challenge at ICRA 2024

Simulation

The organisers provided the simulation setup of 300 pre-generated world of varying difficulties for the benchmarking of the algorithm. We utilized the the global planner from ROS1 move_base and implemented Model Predictive Control (MPC) as our local planner. Here is the formulation for our MPC.

• The green path is the global trajectory
• The short red path is the MPC planned horizon
• The red dots are the raw Lidar scans
• The white squares are the sampled Lidar scans treated as obstacles in the MPC,
• The cyan squares are the obstacles in the local costmap in the blind spot of the Lidar
• The black patches are the constantly updated global costmap.

Our Github

Github Repository

Physical runs at ICRA 2024

We were invited to ICRA 2024 to participate in the final physical runs

BARN 2024 Challenge participants

MLDA Team with Organizers

We managed to get 2nd place the BARN Challenge 2024!

However, I believe that the greatest experience was that I was AT ICRA! I got to meet and talk to brilliant researchers, listen to their presentation and understand their work! I also met with my previous Robotics supervisors in Singapore. On the whole, the trip had been fruitful!

Here is a fun photo with the head of the “Moving Gundam” from the Yokohama Factory.