Team Description: Pumas Logistics League 2023
Affiliation: Bio-Robotics Laboratory, School of Engineering, National Autonomous University of Mexico (UNAM)
Participation: RoboCup Logistics League (RCLL 2023) & Mexican Tournament of Robotics (TMR 2023)
Team Members: Mitzi Ramírez, Rebeca Muñoz, Jesús Savage, Marco Negrete, Julio Martínez, Daniel Ruelas, Miguel Sánchez, Isaac Hernández, Joshua Martínez, and Saul de la Rosa.
1. Background & Overview
With an extensive track record in the RoboCup@Home Open Platform (OPL) and Domestic Standard Platform (DSPL) leagues with Justina and Takeshi, the Pumas Team expands into industrial robotics by competing in the RoboCup Logistics League (RCLL) for the first time.
The objective is translating navigation, decision-making, and sensor-fusion algorithms developed in the laboratory toward dynamic, factory-floor production scenarios involving multiple Modular Production System (MPS) stations.
2. Software Architecture: ViRBot Adaptation
The system adapts the layered ViRBot (Virtual and Real Robot) software framework over ROS to coordinate industrial autonomy:
- Input Layer: Acquires data from LiDAR laser scanners, RGB-D point clouds, magnetic proximity sensors, and infrared sensors to construct consistent symbolic environment states.
- Knowledge Management Layer: 2D mapping using SLAM techniques, Kalman Filter localization, and rule-based expert logic implemented in NASA's CLIPS.
- Planning Layer: Integrates incoming instructions with situation validation, decomposing production line schedules into action sequences.
- Execution Layer: State machines running discrete operational primitives (FindObject, GotoPlace, AlignMPS, AlignObject) with reactive exception handling.
3. Key Technical Innovations
A. RefBox Communication & State Synchronization
Integration with the automated RoboCup Referee Box (RefBox) using Google Protocol Buffers over continuous UDP broadcasts, allowing the robot to receive dynamic game phases, machine states, and factory order assignments under unstable wireless environments.
B. Pumas Navigation Framework
Autonomous navigation tailored for mapped grid areas:
- Global path planning through the A* algorithm.
- Robot pose tracking via AMCL combining wheel odometry and Hokuyo LiDAR scans.
- Real-time 3D point cloud obstacle extraction to avoid dynamic collisions in the factory field.
C. Active Vision, MPS & Gripper Alignment
A multi-stage vision pipeline for piece manipulation and machine docking:
- Station detection using ARUCO marker identification.
- Station plane and boundary fitting to align the mobile base with conveyor belts.
- RGB-D color segmentation and centroid extraction to map 3D piece coordinates via homogeneous coordinate transformations.
- Active-vision camera pose optimization based on utility functions to plan collision-free grasping trajectories.
4. Robot Festino: Technical Specifications
| Mobile Base | Festo Robotino (Omnidirectional base) |
| Vision & Sensing | LiDAR laser scanner, RGB-D camera (point cloud processing) |
| Manipulation & End-Effector | Custom parallel gripper equipped with magnetic proximity and infrared sensors |
| Middleware & Software | ROS, Google Protocol Buffers (RefBox UDP interface), OpenCV, NASA CLIPS, Pumas Navigation |
| Onboard Compute | Dedicated mounted processing laptop connected via Gigabit Ethernet and Wi-Fi |