Move
G1 provides biped movement and arm motion; Go2 brings four-legged gait, turning and traversal across suitable terrain.
Unitree education & research
Robotics for learning and research.
Explore motion, sensing and control with Unitree G1 EDU and Go2 EDU through supervised, measurable projects.
What can a Unitree robot do?
In a supervised lab, learners can compare a planned movement with sensor readings and the robot's observed response. Unitree's mobile hardware and software interfaces make that question testable on a physical platform.
Sense
Use model-specific camera, LiDAR, depth, audio and motion data to observe the robot and the surrounding space.
Navigate
Develop obstacle response, localization and route behaviour for a defined area, with a clear operator fallback.
Interact
Build visual or audio responses. G1 can support object-handling research when fitted with a compatible optional hand.
From perception to action.
Use three connected layers to explain a result: what the robot could sense, which control choice it made, and what its body actually did.
- Execution
- Legs, arms and joints produce movement. Stability, terrain and payload limits shape the physical task.
- Perception
- Sensors provide images, geometry and motion feedback. Coverage and data access depend on the model and installed hardware.
- Decision
- Software turns signals into a control choice. Autonomous behaviour generally requires development, testing and a suitable operating environment.
A route test learners can explain
On a short, supervised Go2 route, the team can compare its planned path with sensor data, movement and operator interventions. The record helps explain where a result diverged from the plan.
- InputRecord the test area and available sensor data.
- InterpretationIdentify the space and obstacles relevant to the route.
- ActionRun the defined movement or stop with operator oversight.
- EvidenceCompare the planned route, observed path and interventions.
Framework adapted from the Unitree manufacturer manual, pp. 3-6. The route example illustrates a developed workflow, not an out-of-box autonomous function. Capabilities vary by model, configuration and software.
Two platforms. Different strengths.
Choose by the physical problem you want to explore: human-scale movement and interaction, or agile four-legged mobility and environmental sensing.
G1 EDU
Human-scale movement and interaction
A biped platform for locomotion, balance, perception and human-robot interaction research. G1 EDU supports secondary development, while hands and additional computing are configuration choices.
- Movement
- Biped locomotion, body balance and articulated arms.
- Perception
- Unitree lists a depth camera, 3D LiDAR and microphone array for the G1 family.
- Extension
- Optional dexterous hand and compute module for suitable manipulation or AI projects.
Best starting point: human-scale motion, interaction and manipulation research when the exact hand and interface requirements are known.
Explore G1 EDU
Go2 EDU
Agile mobility and spatial awareness
A four-legged platform for movement and spatial sensing. Unitree lists 4D LiDAR L2 and a wide-angle camera, with richer development and accessory options in the EDU configuration.
- Movement
- Quadruped gait, turning and traversal over suitable uneven surfaces.
- Perception
- 4D LiDAR L2 and camera data for spatial awareness and obstacle-related tasks.
- Extension
- Mapping, following and navigation workflows depend on the configuration, app and project software.
Best starting point: terrain, sensing and supervised mobile-robot prototypes, including inspection research.
Explore Go2 EDUWhich platform fits the job?
Choose the body around the task first. Then confirm the exact configuration, rather than assuming every robot in a product family has the same development access or accessories.
Reach & handleConsider G1 EDU when the experiment depends on arms, reach or human-shaped movement. Object handling also needs a suitable hand, a defined object range and application-specific control.
Move & observeConsider Go2 EDU when the priority is moving sensors through a test space. Review the route, floor conditions, clearance and required data before selecting accessories.
Access & computeConfirm which interfaces are available on the quoted SKU, what runs onboard and what needs an external computer. Match firmware, SDK and network requirements before development starts.
Whole-project fitAllow for batteries, charging, storage, operator time, spare parts and software work. A robot purchase and a validated application are separate parts of the project.
These are platform directions, not turnkey application guarantees. Confirm the precise SKU, optional hardware, developer access and operating limits before choosing a system. See Unitree's official G1 and Go2 specifications.
What can teams investigate?
Start with a physical question that connects movement, sensing and decisions. These six directions can become supervised learning or research projects when the hardware and software match the task.
Object-handling research
Combine perception, reaching and grasp planning to investigate manipulation. This needs a compatible optional hand, control software and carefully defined objects and safety limits.
G1 EDU with suitable hand. Image: official Unitree dexterous-hand demonstration.
Spatial sensing and mapping
Use Go2's LiDAR and camera to investigate where the robot is, what it sees and how it responds to obstacles. Mapping and autonomous routes require the appropriate software setup.
Go2 EDU. Image: point-cloud view from Unitree's official mapping demonstration.
Motion and balance research
Study gait, body stability and recovery under controlled changes in surface, speed or direction. Compare a planned movement with joint and body feedback.
G1 EDU / Go2 EDUInspection prototypes
Test a repeatable patrol route, capture visual information and review exceptions with an operator. A prototype is not a certified industrial inspection system.
Go2 EDUHuman-robot interaction
Explore how a humanoid detects a person, responds through sound or movement, and communicates its next action. Behaviour depends on the application built around the robot.
G1 EDUSimulation-to-robot validation
Develop a control or perception idea in simulation, then test how it transfers to physical hardware. Unitree publishes SDK and simulation resources; compatibility should be checked for the chosen model.
G1 EDU / Go2 EDUApplication directions use the Unitree manufacturer manual, official G1 and Go2 specifications, and Unitree's open-source resources. Each use case requires a configuration and feasibility review.
Turn a robot demo into a learning experiment.
A controlled challenge gives students or researchers a question, evidence to collect and a result they can explain. These are illustrative investigations, not included courses or ready-made robot functions.
Motion and balance
Change G1's direction or speed in a controlled space. Compare planned movement with body feedback and observe recovery.
Learn to measure stability and repeatability within a safe boundary.Sensing and routes
Give Go2 a known obstacle on a short supervised route. Log sensor data, stops and operator interventions.
Learn how perception, decisions and movement interact.Object interaction
With a compatible optional hand, test one defined G1 grasp across repeated trials. Record successes and failures.
Learn experimental design and the limits of manipulation.The photograph documents Unitree manufacturer examples. The investigations above are suggested test patterns, not completed SpeedyDrone deployments or turnkey features.
Plan a project you can measure.
Start with one learning or research question, the result your team will record, and a safe place to test it. Then choose the robot configuration and control approach.
Test space
Define surfaces, lighting, obstacles, bystanders and a clear boundary for the trial.
Robot setup
Match movement, sensors, compute, battery and optional hardware to the question.
Data & code
Confirm the sensor streams, interfaces and software needed to record and repeat the test.
Supervision
Assign who starts, observes and stops each run, plus charging and equipment checks.
Choose the control approach
Operator-led
A person directs the robot and reviews feedback. Confirm controller, connectivity and line-of-sight needs.
Developed autonomy
Project software handles a defined task within test boundaries, with a human fallback. Confirm sensing, compute and development scope.
Official tools for the development team.
These Unitree resources are starting points. Check support, firmware and setup requirements for the exact model before choosing an interface.
- SDK2
- Robot communication libraries and examples. View SDK2 →
- ROS 2
- Robot messages and examples for compatible models. View ROS 2 →
- MuJoCo
- Simulation for controller development; physical validation remains necessary. View MuJoCo →
- Docs
- Model-specific prerequisites and procedures. Open developer centre →
Three questions to bring to a project discussion
These answers help separate the robot configuration from the software and supervision the project still needs.
OutcomeWhat will learners measure, record or explain after the test?
SetupWhich space, sensors, accessories and interfaces does that result require?
LimitsWho supervises the trial, and what triggers a stop or reset?
Hardware options, developer access and any integration work require confirmation for the exact project. A prototype discussion does not imply certified field readiness.
How institutions use the platform.
The manufacturer manual shows Unitree robots as a base for applied research, interaction and motion-control work. These are Unitree examples, not SpeedyDrone customer cases.
Singapore Institute of Technology
The manual describes a joint-laboratory direction including embodied intelligence, human-robot interaction, safety and applied research.
City University of Macau
The manual highlights quadruped motion control, policy learning and simulation validation as research directions.
Institutional descriptions and photography are attributed to the Unitree manufacturer manual, pp. 21-22. They do not imply a relationship with SpeedyDrone.
Questions before choosing a robot.
For a teaching lab or research project, start with the learning goal, test environment and level of development your team can support.
What is G1 EDU suited to?
G1 EDU is a humanoid development platform for biped movement, perception and human-robot interaction research. Manipulation is possible only with a suitable hand and application software.
What is Go2 EDU suited to?
Go2 EDU is a quadruped platform for terrain movement, spatial sensing and mobile-robot development. It is a useful base for mapping, obstacle and route-planning prototypes when the hardware and software match the task.
Can G1 EDU pick up objects?
A compatible dexterous hand is optional on G1 EDU. Reaching and grasping also require object-specific perception, control development and safe testing; the base robot should not be assumed to perform a complete manipulation task out of the box.
Can Go2 EDU map and navigate autonomously?
Unitree describes LiDAR mapping and app-based path functions for Go2. The available workflow depends on the exact configuration, software and test environment. A project-specific autonomous task requires validation and human oversight.
What development resources are available?
Unitree publishes SDK, ROS and simulation resources for supported robots. Confirm compatibility, data access and the required developer interfaces for the exact model and software version before planning an integration.
How can an education team use a Unitree EDU robot?
A school or research team can design supervised investigations in motion, sensing, interaction or control. Define one question, the required hardware and software, a safe test boundary and a way to review results. Courses and applications are not included with the robot unless explicitly specified.
How do I choose between G1 EDU and Go2 EDU?
Choose G1 when human-scale movement or interaction is central. Choose Go2 when four-legged mobility and spatial sensing are central. Then confirm the sensors, options, interfaces, operating conditions and test criteria for your task.
What could your team learn or test?
Tell us whether you are planning a teaching lab, student project or research prototype. Share the task, environment and required interfaces so we can discuss a relevant Unitree configuration.
Manufacturer examples and documentary photographs come from the Unitree manual and official Unitree product pages. Platform product photographs come from current SpeedyDrone listings. Features, options and development resources vary by exact configuration. See Unitree's G1, Go2 and open-source information.