A repeatable companion robot test protocol
A practical protocol for measuring setup, sensing, movement, conversation, privacy controls, offline behavior, noise, charging, and failure recovery.
Products, prices, policies, and evidence change. Verify the current primary sources for the exact model and region before acting.
Freeze the test conditions
Record model, firmware, app version, plan, language, room, lighting, network, floor type, battery state, and date. Use the same prompts and tasks for comparable robots while allowing product-specific interactions to be documented separately.
Separate a vendor claim, a successful observation, and a repeatable result. One charming demonstration is evidence that something happened once, not that it will work reliably in every home.
Was the task repeated?
Were failures recorded?
Could an observer reproduce the setup?
Did updates change the result?
Test normal use and failure
Measure setup time, wake reliability, recognition under different light and noise, navigation around realistic obstacles, return-to-charge success, false activations, physical mute behavior, app permissions, account recovery, and graceful behavior during Wi-Fi loss.
For conversation, use the same factual, ambiguous, emotional, refusal, and safety prompts. Report latency, interruptions, uncertainty language, escalation, and whether the device invents capabilities or professional authority.
Publish limitations and raw observations
State sample size, test duration, funding, supplied hardware, conflicts, regional differences, and anything you could not inspect. Share a redacted observation sheet and keep private household data out of recordings and logs.
Framework sources
These sources shape the questions in this guide. They do not certify any listed robot.