UruBots — RoboCup@Home Open Platform League 2027

Team UruBots · Universidad Tecnológica del Uruguay (UTEC) — ITR Norte, Rivera, Uruguay · Target event: RoboCup 2027, Nuremberg, Germany, June 15–21, 2027.

Qualification Materials (RoboCup@Home OPL 2027)

Base@Home platform

Base@Home is a custom low-cost domestic service robot developed by UruBots for the RoboCup@Home Open Platform League (OPL), at Universidad Tecnológica del Uruguay (UTEC), ITR Norte, Rivera.

Mobile Base & Navigation

Differential-drive base built with dual 6.5" BLDC hoverboard hub motors (custom STM32F103 FOC firmware, 90 Hall ticks/rev, track_width = 0.5620 m, wheel_radius = 0.0853 m) and a Hokuyo UST-05LX 2D LiDAR (270°, 0.06–5 m, 40 Hz, beam height 38 cm), running Nav2 + AMCL + SmacPlanner2D + RegulatedPurePursuit on ROS 2 Jazzy.

Torso & Manipulation

NEMA23 / ESP32-S3 vertical prismatic lift column and a side-mounted 6-DOF AgileX Piper manipulator with parallel gripper and an eye-in-hand Orbbec DaBai DC1 camera, planned with MoveIt 2 (TRAC-IK + OMPL) and supervised via a FollowJointTrajectory action server.

Head & HRI

2-DOF Dynamixel XL430-W250 pan-tilt head carrying an Intel RealSense D435 RGB-D camera, a RODE VideoMic GO II directional shotgun microphone, an animated facial display and a TTS speaker.

Distributed Compute

Wired 1 Gb/s Ethernet, ROS 2 Jazzy + CycloneDDS unicast, no team Wi-Fi hotspots. Host 1 (Intel NUC i3-1220P, Ubuntu 24.04): Nav2, AMCL, RealSense D435 face/pose perception (YuNet/SFace, YOLOv8-pose), Receptionist & Mission Executor, deterministic bilingual NL/GPSR compiler and a local URDF/Xacro web editor. Host 2 (NVIDIA Jetson Orin Nano Super 8 GB, JetPack 7.2.1 / L4T R39.2.1, Ubuntu 24.04): 50 Hz CAN Piper controller, U2D2 pan-tilt node, Orbbec DaBai wrist stream + TensorRT FP16 YOLO detector (55 FPS end-to-end on GPU vs. 5.2 FPS on CPU, verified on hardware), and on-demand faster-whisper ASR with Silero VAD.

Safety & Evidence Note

All actuator drivers boot disabled by default (auto_enable=False on Piper, auto_torque=false on pan-tilt), reject motion commands without fresh hardware feedback, and never publish synthetic zeros as measured joint states. Capabilities validated only in simulation (SIM_PASS, such as the natural-language Commander and trash-disposal sequence) validate software logic and kinematics only and do not demonstrate physical safety on real hardware.

Open-source software

Packages contributed by the team:

  • hoverboard_base
  • base_athome_bringup
  • base_athome_perception
  • base_athome_moveit_config
  • base_athome_skills
  • base_athome_sim
  • jetson_head
  • tools/urdf_editor
  • tools/gpsr_parser