TRON1 Education Edition · LimX Dynamics × Collie

Teach embodied AI on a real robot, not just in a simulator.

TRON1 is a multi-modal biped robot with an open SDK, a 12-module curriculum and a complete sim-to-real pipeline. Collie delivers it with on-site faculty certification and ongoing local support, so your lab is teaching from week one.

verifiedOfficial LimX EDU Partner schoolSyllabus Ready support_agentSingapore HQ Engineering Support
TRON1 wheeled-leg robot, 45 degree view
3-in-1
Point-foot, sole-foot & wheel-foot locomotion
≤ 20 kg
Lab-friendly size, one person can handle it
12 modules
From robot basics to RL, vision, voice & SLAM
C++ · Python
Open SDK with ROS & ROS 2 support
2 days
To certify your faculty as TRON1 trainers
Why TRON1 for education

One platform that grows with your programme

Students learn the same tools industry uses, on hardware they can actually run their own code on, with safety and assessment designed in from the start.

security

Real hardware, safe to learn on

Every new program runs simulation first, then on a gantry, then on the floor. Physical and remote e-stops, a named operator and spotter, and a session log keep the lab safe.

Safety SOP enforced
terminal

Open-source, industry toolchain

Official LimX repositories cover the full train-to-deploy path, alongside NVIDIA Isaac Lab, MuJoCo, Gazebo and ROS 2, so graduates take skills straight into industry.

Isaac Lab + ROS 2 ready
swap_horiz

Three ways to move

Swap between point-foot, sole-foot and wheel-foot ends to compare locomotion strategies (walking, balancing and rolling) on a single robot chassis.

Fast hardware swapping
extension

Expandable for capstones

Add the arm, sensor/LiDAR and voice-interaction kits when your programme is ready for mobile manipulation, SLAM and language-driven embodied AI agents.

Manipulator & LiDAR ready
The sim-to-real learning path

From a policy in simulation to a robot walking in your lab

01 · Simulate

Gazebo & ROS

Run sample joint-move programs and watch kinematics in real time in PlotJuggler or RViz.

02 · Train

NVIDIA Isaac Lab

Train locomotion policies with reinforcement learning across thousands of parallel environments.

03 · Cross-check

MuJoCo

A policy that works in two physics engines is far more likely to work reliably on hardware.

04 · Gantry

Suspended first run

Deploy with Python or ROS 2 while the robot hangs safely, feet just millimetres off the floor.

05 · Floor

Real-world run

Run in a marked test zone with a spotter, then review the rosbag and log the session.

info Optimal lab economics. About 80% of lab time happens in simulation, so a single robot can serve many student teams, each one booking real-robot time only after their code passes.
Curriculum architecture

12 modules, three tiers, ready to drop into your syllabus

Mapped directly to the official LimX curriculum and structured so students progress systematically from first power-on to an integrated capstone application.

Tier 1 · Modules 1–5 Base unit

Foundations

For introductory and second-year engineering students

  • check_circleTRON1 introduction and configurations
  • check_circleROS 2 and the LimX SDK setup
  • check_circleAI fundamentals in robotics
  • check_circleSensors: IMU and foot/force telemetry
  • check_circleBasic motion control & inverse kinematics
Prerequisites: Python or C++ basic literacy
Tier 2 · Modules 6–9 Base unit + GPU PC

Intelligent Control

For senior undergraduates and capstone teams

  • check_circleAI agents: perceive, decide, act pipelines
  • check_circleModel Predictive Control (MPC) vs. Reinforcement Learning
  • check_circleAI vision integration and dataset fine-tuning
  • check_circleHands-on RL policy training & deployment
Isaac Lab + MuJoCo pipeline
Tier 3 · Modules 10–12 Expansion kits

Capstone

Optional, for advanced master's and capstone research

  • check_circleVoice interaction with on-campus LLMs (Ollama)
  • check_circleMobile manipulation using TRON1 arm kit
  • check_circleSLAM & 3D autonomous point cloud navigation
Full system integration & thesis

Assessment-ready from day one

Each tier comes with auditable learning outcomes, evidence requirements and a four-level rubric, so university markers grade practical lab sessions and capstone defences consistently.

Safety discipline Technical execution Method & rigour Evidence & communication Collaboration & handover
Safety built in

A five-gate sign-off before any new code touches the floor

Prevent laboratory equipment collisions and guarantee safe undergraduate exploration with Collie's certified operational sign-off sequence.

G1

Simulation pass

Runs to completion in virtual sim, results fully recorded.

Prerequisite
G2

Second-engine check

Cross-verified in MuJoCo physics engine before compile.

Verification
G3

Gantry run

Suspended on lab tether, remote physical e-stop verified.

Tethered test
G4

Floor run

Zone perimeter cleared, designated spotter on duty.

Live physical
G5

Post-run record

Full rosbag saved to lab server, session log certified.

Audit & grade
Open-source toolkit

The tools industry already uses

Every graduate trains on battle-tested frameworks directly transferable to enterprise robotics ventures and leading research institutions.

ROS 2 NVIDIA Isaac Lab MuJoCo Gazebo rsl_rl Pinocchio · OCS2 OpenCV Ultralytics YOLO Nav2 · FAST-LIO2 faster-whisper · Ollama PlotJuggler · RViz · Foxglove Docker · GitHub Classroom
lock_person Local data governance. Offline voice (faster-whisper) and language models (Ollama) keep student speech data completely on-campus, simplifying institutional ethics and IT approval.
Engineering benchmark

Key specifications

Dimensions≤ 392 × 420 × 845 mm
Net weight≤ 20 kg
Payload≤ 10 kg
Speed< 1 m/s legged · ≤ 3 m/s wheeled
Terrain adaptability≤ 30° slopes · ≤ 20 cm obstacles
BatteryUp to 2 h · quick swap · < 1 h fast charge (20–80%)
Onboard compute12th Gen Intel i3 · 16 GB · 512 GB
Actuators30 N·m rated · 80 N·m peak torque
Expansion portsUSB 3.0 · GbE · 24 V / 100 W power port
Remote controlHandheld controller, 50 m range, remote e-stop

* Specifications as certified by LimX Dynamics; measured under standard laboratory conditions.

Train-the-trainer

Your faculty, certified in two days

We certify 2–4 of your lecturers and lab technicians first, so your team runs student modules completely independently, with Collie specialists one call away.

verified_user
17-point competency checklist Every trainer executes the complete sim-to-real flow solo to receive accreditation.
layers
One standard Docker image Pre-configured with ROS, LimX SDK and simulators, ensuring every workstation is synchronised.
forum
Monthly 1-hour faculty clinic Direct hotline to Collie senior robotics engineers; firmware issues escalate directly to LimX.
Day 1 · 09:00 to 17:00

Hardware & simulation

  • Safety SOP and e-stop operational drill
  • Power-on, calibration & operating modes
  • Network setup & on-robot web utilities
  • Developer PC and workspace initialisation
  • Gazebo simulation, PlotJuggler, RViz
  • First program on the real robot (gantry)
Phase 1 milestone: gantry sign-off
Day 2 · 09:00 to 17:00

Learning & deployment

  • RL locomotion training in Isaac Lab
  • Policy cross-check inside MuJoCo
  • Deploying trained policy: gantry to floor
  • Hardware troubleshooting protocols
  • Telemetry logging & rosbag diagnostic review
  • Final faculty solo certification run
Phase 2 milestone: official accreditation
See it in action

TRON1 in motion

Short real-world clips of TRON1 on the move, the kind of behaviour your students will be programming.

Institutional assurance

Frequently asked questions

Yes. About 80% of lab coursework happens in simulation directly on students’ workstations. Student teams book 30-minute real-robot test slots only after their code passes Gates 1 and 2 in simulation. A single TRON1 chassis easily supports 25 to 40 students per semester.

Standard student laptops or lab desktops with Ubuntu 22.04 LTS, 16 GB RAM and 512 GB storage are sufficient for SDK, ROS 2 and Gazebo modules. For Tier 2 reinforcement learning policy training, your department requires at least one dedicated workstation equipped with an NVIDIA RTX GPU (12 GB+ VRAM) running NVIDIA Isaac Sim and Isaac Lab.

No prior kinematics or physical robotics experience is needed. Tier 1 starts from first power-on, Linux fundamentals and basic ROS 2 publish-subscribe loops. Tiers 2 and 3 build systematically toward advanced reinforcement learning, sensor fusion, computer vision, local LLM inference and SLAM.

Collie provides official global university distribution with certified customs handling across Southeast Asia, North America, Europe, Australia and the Middle East. On-site faculty training is conducted by bilingual robotics engineers, backed by rapid hardware swap logistics from Singapore and regional hubs.

Bring TRON1 into your classroom

Tell us about your lab and we’ll recommend the right setup.

Tell Us Your Needs arrow_forward