Compare security robot pilots and reported operations by robot, operating mode, evidence date and cost scope. Inspect Caltrans, NYPD and stadium records and export the sources.
DROID RESEARCH · SOURCES CHECKED
Separate the machine, the operator and the outcome.
These records describe particular sites and periods. Completed pilot, reported operation and operating mode are separate fields. A research report may evaluate a remotely driven system; a current product page may advertise automatic routes. Neither should silently replace the other.
This initial set is not a market-wide ranking or a list of Droid installations. Downloading retains dates, conditions, unknowns and primary sources.
Equipment-yard patrol using two DroneDogs. UC Davis AHMCT reports human RSOC operators piloted each patrol. This study does not validate the autonomy of today’s offering.
Limits of the record
The report identifies weather pauses and missed patrols. Table 4.1 excludes April, boundary periods and maintenance downtime; its averages are not whole-contract uptime.
Reported outcome
No theft or damage incidents were reported during the pilot. Without a comparable control, that observation does not establish a causal reduction.
Price evidence
About $142,000 USD · 2 robots · 6 months · 112 scheduled hours/week · 2024 pilot, San Bernardino, California, US. Approximate total pilot deployment cost. Current quotes, detailed inclusions and longer-term rates may differ.
Droid’s procurement implication
Require two denominators: service hours purchased and useful route coverage delivered. Preserve all downtime in the acceptance log.
Actual patrol time across all scheduled hours, including excluded periods
Current delivered autonomy and the cost of retained human response
Evidence type: commissioned field study · Source date: 2025-05-20 · Reviewed 2026-11-10
Knightscope K5 ↗ · K5 used in 2023–24; exact revision not established
Task and operating evidence
A predefined station route during midnight–6 a.m. hours. NYC’s inspector general describes the concluded pilot and a deployment plan with TARU personnel supervising. The launch called for an accompanying officer.
Limits of the record
This is a historical pilot record. The report does not establish the current status of the unit or performance of the K5V5 brochure configuration.
Reported outcome
The inspector general reported that requested outcome records had not been supplied at publication. Effectiveness remains unestablished in this source.
Price evidence
No current configured price established in this record.
Droid’s procurement implication
An escorted trial does not demonstrate removal of a staffed post. Measure retained labor rather than assuming it disappears.
Pilot outcome records and actual staff time
Any later redeployment and its exact configuration
After-hours mobile video patrols, reported by the stadium itself. The stadium identifies BENZIE in its security operation. Asylon’s interview links the deployment to its platform; its case study describes remote alert review.
Limits of the record
The reviewed customer page does not quantify autonomy, patrol availability, staffing hours or operating duration. Separate reports are not continuous telemetry.
Reported outcome
Asylon’s case study claims 2,000+ patrols, zero theft/vandalism in the monitored zone and 30% cost savings. The page does not provide a dated observation window, cost denominator or counterfactual; these are supplier claims, not audited effects.
Price evidence
No current configured price established in this record.
Droid’s procurement implication
A named customer establishes more than a demonstration. It still does not establish your site’s economics or an independently measured service level.
What exact cost base and observation period produced the claimed savings?
What share of patrol execution requires remote driving?
Evidence type: customer and supplier reports · Source date: Mixed dated and undated sources · Reviewed 2026-11-10
Log every scheduled minute as useful patrol, charging, weather hold, maintenance, blocked route or another explicit state. Report the longest checkpoint revisit gap as well as the average. Two robots moving together do not necessarily cover twice the required route.
Measure retained human work.
Separate remote driving, mission dispatch, alert review, on-site response and repairs. Record which hours change an actual budget. More information for an existing team can be valuable without implying a removed guard position.
These are Droid’s proposed measurement rules. They are not a claim that every reviewed vendor already reports these metrics.