DATA CENTER DEPLOYMENTS
Define the workflow.
Set the acceptance criteria.
Scope a supervised humanoid trial around a rack list, a task protocol, and an agreed scorecard. Include failed attempts, recovery time, and operator control minutes.
Specify the pilot01 / CHOOSE A STARTING WORKFLOW
Select the task protocol.
Proposed pilot scopes. Hardware, sensing, routes, and operator responsibilities require site qualification.
Rack inspection and asset audit.
Register the rack list, U positions, inspection points, and expected asset identities before the round.
- Read asset tags and serials; compare a sampled set with hand-checked ground truth.
- Check visible indicators, cable obstructions, and the site-approved inspection checklist.
- Attach rack/U IDs, timestamps, and images to findings; site staff confirm or reject each anomaly.
Racks inspected / scheduled; exact tag matches / sampled tags; confirmed, rejected, and unresolved findings; operator minutes per completed task.
Evaluation protocolField pilot / Measurement protocol
Nine operational measures.
Agree the task, denominator, operating window and acceptance threshold before the first run. These are proposed pilot definitions; measured results are not published here.
| Metric | Definition | Required reporting |
|---|---|---|
| 01Uptime / robot availability | Available robot hours ÷ scheduled robot hours. | Fix the readiness criteria and schedule before the trial. Break out charging, repair, software faults and network downtime within that window. This measures the robot, not facility uptime. |
| 02Round completion | Scheduled racks inspected with required evidence ÷ racks scheduled, per round. | Retain the scheduled rack list. Record completed, blocked and skipped racks with reason codes; a repeat visit does not add another completed rack. |
| 03Findings | Anomalies raised, confirmed by site staff, rejected as false positives, and pending review. | Attach rack/asset ID, timestamp and image evidence. Retain the site adjudicator and disposition. A raised alert is not a confirmed fault. |
| 04Audit accuracy | Exact asset-tag matches ÷ tags in a hand-checked sample. | Register the sample before scoring. Include unread and misread tags in the denominator; report each separately. Reconcile rack position and serial number where in scope. |
| 05Parts run time | Elapsed time from request receipt to verified part handoff at the target rack. | Include dispatch/acceptance delay, queue, pickup, transit and handoff time. Report p50/p95 for completed runs alongside outstanding, aborted and failed requests. |
| 06Intervention rate | Task attempts with an operator takeover ÷ all initiated attempts. | Keep aborted and failed attempts. Log takeover reason, control ownership and recovery outcome. Report planned manual steps separately from unplanned interventions. |
| 07Operator minutes per task | Total operator control minutes ÷ completed tasks. | Sum all operators, failed-attempt control time and recovery within the reporting window. This is a control-time cost driver; report passive supervision and reset labor separately. Zero completions means undefined, not zero. |
| 08Incidents | Equipment/person contacts, near misses, person-triggered stops and emergency stops. | Separate event categories and severity. Retain task, location, recovery and disposition, with operating hours and attempt counts as exposure. Review stops individually. |
| 09Network health | Robot ↔ operator RTT p50/p95/p99, jitter, packet loss, failovers and degraded-mode minutes. | Declare endpoints, probe cadence, sample count and jitter method. Report sent/lost probes separately from delivered-probe latency; tie outages and failovers to mission IDs. Set thresholds with the site. |
03 / SITE QUALIFICATION BRIEF
Specify the operating
envelope.
Document the constraints before a site trial. Attach the route plan, task checklist, baseline timings, and the sample used to verify observations.
Thresholds agreed before the first measured run.- 01
Route and hardware.
Rack/U locations, aisle widths, turning clearance, door thresholds, floor limits, access permissions, and docking space. State payload dimensions, mass, retention, and ESD requirements.
- 02
Network and stop behavior.
Identify the robot-to-operator path. Measure RTT p50/p95/p99, defined jitter, loss, roaming, and failover. Qualify local stop behavior, degraded operation, and recovery when connectivity fails.
- 03
Operator and evidence.
Define who can approve, take over, stop, and restart each task. Specify work-order IDs, handoff acknowledgement, access to images/logs, and retention periods. Agree a hand-checked sample, denominators, thresholds, and stop conditions.
04 / TRIAL AND ACCEPTANCE
Pre-register the protocol.
Freeze the scheduled rack list, task definition, baselines, sampling method, operating limits, and acceptance thresholds. Assign site and operator responsibilities.
Execute and review.
Log every attempt, including skips, failures, takeovers, recovery, stop events, and network degradation. Site staff adjudicate findings against timestamped evidence.
Accept, hold, or expand.
Compare the complete run set with the agreed criteria. Hold unresolved failure modes; qualify any change to route, payload, hardware, or autonomy before expanding scope.