Unitree G1 Repair: Fall, Joint and Calibration Guide

Robotics repair branch

A Unitree G1 repair decision begins with the exact robot, the exact event and the operator's required outcome. A fall, balance warning, joint alarm or uneven step does not identify one failed part by itself. The robot's physical condition, supported status information, environment and repeatable behavior must stay connected as one evidence file. This guide keeps the customer's concern intact while separating safe observation from work that belongs in a controlled robotics workshop.

Quick answer

Stabilize the G1, preserve the first evidence and avoid turning one symptom into a parts verdict

After a fall or abnormal movement, stop the robot in a clear area, remove people from its reach and use an appropriate support frame before any powered observation that a qualified team approves. Photograph the exact stance, contact marks, joint housings, feet, supplied kit and warning state. Review Unitree Explore status, temperature and alarm information only as evidence, not as a fixed diagnosis. Reboot Hub then documents known and unknown findings, proposed scope, parts provenance and a controlled acceptance plan around the customer's real task.

What evidence should be captured before the customer acts?

Evidence layer	What to record	Why it changes the decision
Exact unit identity	G1 label, configuration, supplied controller, charger, battery and accessories	Prevents evidence from another G1 configuration or kit shaping the decision
Event record	Last normal movement, fall direction, surface, load, warning, sound and operator action	Preserves the sequence before resets or repeated standing attempts change it
Physical evidence	Exterior panels, joint housings, feet, fasteners, gaps, scuffs and stance	Shows where impact or load evidence exists without claiming an internal failure
Supported status	Unitree Explore status, temperature, alarm and calibration information available for the exact robot	Keeps software-visible evidence tied to the physical unit and time
Required outcome	Walking surface, payload, interaction boundary, runtime, environment and deadline	Defines acceptance around what the customer actually needs the robot to do

How should common Unitree G1 symptoms be separated?

Reported concern	Safe evidence to capture	Decision boundary
Will not stand	Record stance, support, battery state, app status and the last event	Do not command repeated standing attempts until structural and powered risk is bounded
One leg moves differently	Capture side-by-side movement, sound, joint exterior and alarm timing	Asymmetry can have mechanical, calibration, power, control or event-related causes
Falls after a few steps	Record surface, direction, gait mode, temperature, load and repeat point	A fall pattern is evidence; it is not a single-actuator diagnosis
Joint alarm or abnormal heat	Preserve exact warning, joint location, ambient condition and idle behavior	Stop when heat, odor, binding or uncontrolled movement makes continued testing unsafe
App or connection concern	Record connection stage, device, network context and robot-side status	Separates access to evidence from locomotion, battery and sensor health
Post-impact body damage	Photograph panels, gaps, fasteners, feet and joint housings before moving them	Keeps visible impact evidence available for a bounded workshop scope

How do you confirm the exact Unitree G1 before diagnosis?

Begin with the physical robot rather than an order screenshot or a remembered configuration. Photograph the product label, overall body, hands or end effectors if fitted, joint housings, feet, battery interface, charger, controller and every accessory supplied for diagnosis. Record which parts belong to this robot and which were borrowed from another kit. If a customer operates more than one G1, assign an evidence label before files, alarms or photographs are collected so the record cannot drift between units.

Configuration matters because a G1 can be used for different demonstrations, research tasks and development environments. A symptom description such as arm error or unstable walk is incomplete without the affected side, operating mode, surface, payload, recent event and software-visible state. The goal is not to infer a hidden component from a model name. It is to preserve a traceable exact-unit baseline that a technician and customer can both review before scope is approved.

Unitree G1 powered off on padded mat while technicians document fall evidence — Preserve the fall position, supplied kit and contact evidence before commands or calibration change the baseline.

What should happen immediately after a fall or uncontrolled movement?

Clear people and loose objects from the robot's reach, stop further commands and preserve the position if it can be done safely. Do not pull on an arm, leg, hand or exposed cable to reposition the robot. Use a support frame and a trained handling plan when the robot must be stabilized or observed under power. Photograph the fall direction, contact points, floor surface, nearby obstacles and any load that was attached before evidence is disturbed.

A robot that looks intact may still have an unknown mechanical, sensing or control boundary. Repeated attempts to stand can add impact and erase the first failure state. Record odor, heat, noise, visible gaps, looseness and warning behavior without forcing motion. If the battery, enclosure, joint or powered movement appears unsafe, stop. The first decision is whether controlled inspection can continue, not which replacement part should be ordered.

Which arrival evidence makes a G1 repair quote more reliable?

The intake record should include full-body photographs from several angles, close views of each reported area, the supplied kit, packaging condition and the customer's account in their own words. Record the last normal task, when the symptom began, any fall or collision, prior service, transport event, update, calibration attempt and environmental change. Keep original video and app screenshots linked to the exact date and unit rather than compressing them into a message thread with no context.

Ask what result the customer needs after service. A robot used for a controlled demonstration has a different acceptance boundary from one expected to repeat a research gait on a particular surface. The quote becomes stronger when it names the use case, unknowns, diagnostic limits and evidence still needed. It becomes weaker when it promises a component before the incident timeline and exact-unit baseline are preserved.

How should Unitree Explore information be used without overdiagnosing?

Unitree Explore can provide supported status, temperature, alarm and robot-management information for the exact G1. Capture the visible state before clearing warnings, changing settings or attempting calibration. Record the connected device, time, robot identity and the condition under which the information appeared. A screenshot is useful only when it stays tied to the physical evidence and operator report; an isolated code or color is not a complete diagnosis.

Use the app to organize questions, not to skip physical inspection. A reported joint warning may align with impact evidence, abnormal resistance, temperature or a repeatable movement pattern, but those relationships still require professional interpretation. If the app cannot connect, preserve that as a separate symptom. Do not assume the connection concern caused the movement concern or that restoring access proves the robot is mechanically safe.

Unitree G1 status and alarm evidence reviewed safely in Unitree Explore — Supported app information is one evidence layer; it does not replace physical inspection or professional interpretation.

When do auxiliary calibration and joint fine-tuning belong in the process?

Auxiliary calibration and joint fine-tuning are supported tools for trained work on a correctly identified and physically safe G1. They should follow, not replace, evidence capture and mechanical-risk screening. Record the as-received stance, app information and movement pattern before changing calibration-related values. If impact, binding, structural damage, abnormal heat or uncontrolled motion is present, a calibration attempt can hide the baseline or command movement that should not occur.

The approval record should explain why a calibration step is proposed, what prerequisite checks passed and how its result will be evaluated. A temporary change in stance does not prove that an underlying joint, sensor, power or structure concern is resolved. Preserve before-and-after evidence and return to a controlled support frame when movement could be unpredictable. Never present calibration as a universal fix for a robot that falls.

How do you document uneven gait without naming a failed actuator?

Use a clear, level, low-risk test area only after the robot has passed the safety boundary. Record front, side and rear video at normal speed, then note which foot, joint or phase differs and whether the pattern repeats. Keep the gait mode, speed, surface, load, battery state, temperature and support conditions constant. Change one variable at a time so the evidence remains interpretable and does not turn into a collection of unrelated clips.

Uneven gait can reflect more than one layer: physical resistance, alignment, calibration, sensing, control, power delivery, surface interaction or consequences of a previous fall. The technician should document observed behavior and the tests that narrow it, while leaving unknowns visible. A customer can then approve a bounded diagnostic scope instead of being told that one actuator is defective solely because one leg appeared late in a video.

What can joint noise, resistance and exterior marks actually prove?

Joint noise and resistance become useful when their location, timing and operating state are repeatable. Record whether the sound occurs at rest, during supported movement, under load or only after warming. Photograph joint housings, seams, fasteners and nearby panels under broad light. Do not force a joint through resistance or use a comparison robot as permission to exceed the exact unit's safe range.

An exterior scuff can show contact but cannot by itself prove internal damage. Likewise, a clean housing does not rule out a deeper concern. Keep the visible evidence, supported status and movement record together. If a joint is hot, binding, loose, mechanically displaced or associated with uncontrolled motion, stop powered testing and move the decision to professional inspection rather than repeating the symptom for a better video.

Unitree G1 leg joint housings inspected externally in a support frame — A support frame keeps observation controlled when joint behavior or balance may be unpredictable.

How should battery and power symptoms be separated from movement symptoms?

Record battery identity, charge state, storage history, charger, connection behavior, idle stability and when the symptom appears. A robot that shuts down during movement may have a different evidence path from one that never completes startup. Do not treat a percentage display as proof of battery health, and do not keep cycling a pack that is swollen, damaged, unusually hot, odorous or associated with unstable power.

Power evidence must remain tied to the task. Note whether the concern appears during standing, walking, arm movement, network use or a particular load. A known-good comparison may help a trained workshop, but every substitution must be recorded so a successful test is not misrepresented as proof that the original battery or every downstream layer is healthy. Unsafe packs stay outside owner-side experimentation.

When are sensing, network and locomotion separate branches?

A G1 may be physically able to stand while app access, perception input or network communication is unavailable. Conversely, a normal connection does not prove balance, joint or power health. Record the exact stage that fails: discovery, connection, command receipt, status reporting, environmental perception or movement execution. Keep the device, network, distance, obstacles and robot mode consistent during any safe comparison.

Do not collapse these branches into robot offline. Preserve robot-side indicators, supported status, operator command and visible response in the same timeline. If the robot's movement is unsafe, do not continue merely to investigate a connection symptom. A professional scope can then separate access restoration, sensor evidence and locomotion diagnosis while still showing the customer how the layers interact.

What does a transparent G1 repair decision include?

A transparent decision starts with the customer's concern, exact-unit photographs, event timeline, supported status and reproduced behavior. It states known and unknown findings, the inspection boundary, proposed work, parts provenance, timing and acceptance plan before approval. It does not substitute a list of possible parts for diagnosis, and it does not imply that a fall, warning or uneven step guarantees one internal cause.

Reboot Hub records evidence at intake, quote, approved work and handover so the customer can see what changed and what remains outside scope. If a part or complete unit is used for comparison, that substitution is named. If the required task cannot be accepted safely, the written result should say so. Clear boundaries are part of the service, not a disclaimer hidden after payment.

Unitree G1 completing a controlled post-repair walking acceptance sequence — Acceptance progresses from safe stationary checks to a representative low-risk task, with stop conditions recorded.

When is a documented replacement stronger than G1 repair?

Repair is stronger when the concern is repeatable, risk can be bounded, parts provenance is defensible and the exact robot can pass a controlled acceptance sequence for the required task. Replacement is stronger when impact or powered risk cannot be bounded, multiple systems are affected, critical parts are uncertain, prior modification obscures the baseline or the required outcome cannot be demonstrated within the customer's operational window.

Compare a replacement as an evidence path rather than an automatic upsell. Name the exact configuration, condition, supplied equipment, warranty boundary and remaining unknowns. The decision should account for downtime, deployment purpose and the customer's tolerance for uncertainty. A lower initial repair scope can still be the weaker choice if it cannot produce a defensible acceptance record.

How should controlled acceptance and handover be recorded?

Controlled acceptance moves in stages: safe startup, supported status review, stationary joint behavior, supported movement, low-speed stance and gait, then the customer's representative task only when each prior stage is acceptable. Record the surface, mode, load, duration and observer positions. Stop if heat, noise, alarms, instability or uncontrolled movement returns. A quick power-on is not a completed robotics acceptance test.

The handover should identify the original concern, written findings, approved work, parts provenance, test evidence, residual risks and care boundaries. Reboot Hub's normal workshop target is 1-3 business days after quote approval when diagnosis, parts and acceptance allow. Approved work carries a 30-day repair warranty; eligible pre-owned complete products follow a separate 180-day product warranty. Those terms remain distinct in the customer's written file.

Written service boundary

What Reboot Hub commits to before work begins

Timing

The normal workshop target is 1-3 business days after quote approval, subject to diagnosis, parts and acceptance testing.

Repair terms

Approved repair work has a 30-day repair warranty under the written repair scope.

Product terms

Eligible pre-owned complete products have a separate 180-day product warranty.

Related Reboot Hub paths

Move from evidence to a documented next step

Continue the learning path, submit the exact-unit concern or compare a documented replacement.

Path	Use it for
Support & Learning	Return to the complete structured learning path.
Complete repair path	Follow concern, evidence, diagnosis, approval and handover.
Start a repair ticket	Send the exact-unit concern and available evidence.
Warranty policy	Keep repair and eligible product warranty terms separate.
The Reboot Hub Standard	See how condition, unknowns, evidence and written terms are handled.
Pre-owned Unitree G1	Compare a documented complete-robot path when replacement is stronger.

Unitree G1 FAQ

Questions to settle before approving repair or replacement

Does a Unitree G1 fall prove that one actuator failed?

No. Preserve the fall, physical evidence, supported status, surface, mode and repeatable movement before a technician narrows mechanical, sensing, calibration, control or power branches.

Should I make a G1 stand repeatedly after it falls?

No. Clear the area, stabilize it safely and stop repeated commands until structural, battery, joint and uncontrolled-movement risks are bounded.

Can Unitree Explore diagnose a G1 by itself?

It can provide useful status, temperature, alarm and calibration evidence, but that evidence must remain tied to the exact robot, event and physical inspection.

Is auxiliary calibration the first step for uneven G1 gait?

No. Preserve the original behavior and screen for impact, binding, heat, structural damage and unsafe motion before trained calibration work is considered.

What should a Unitree G1 repair quote include?

It should identify the exact robot, concern, event evidence, known and unknown findings, diagnostic boundary, proposed scope, parts provenance, timing, terms and acceptance plan.

When is replacement stronger than G1 repair?

Replacement is stronger when risk or parts cannot be bounded, several systems are affected, prior changes obscure the baseline or the required task cannot be accepted safely.