Autonomous robots are leaving controlled showrooms and entering warehouses, factories, energy facilities, security operations and public spaces. The equipment is becoming more capable, but the quality of purchasing decisions still depends on the same discipline: verifying claims against the intended operating environment.
A brochure can state speed, payload, environmental protection and navigation capability. It rarely proves what happens when the system is carrying its real sensor suite, communicating with customer software, dealing with blocked routes or recovering from faults. The right procurement question is therefore not “What can the robot do in a demonstration?” but “What evidence shows that it will work safely and predictably every day?”
01 / Define the exact configuration
Start with a configuration baseline, not a model name. The same robot platform may be marketed with several batteries, wheels, sensor options, compute modules and communication technologies. Request a version-controlled specification covering hardware, firmware, accessories, installed payload and intended duty cycle. Cross-check values across the datasheet, manual, offer and actual test report. Any discrepancy should be registered as an open technical question.
For an autonomous quadruped or hybrid wheeled robot, the payload and power consumption of LiDAR, cameras, RTK, 5G, edge compute and continuous streaming can materially change usable endurance. Request data from a representative mission instead of treating an unloaded runtime as an application guarantee.
02 / Test autonomy, not the word autonomous
Ask exactly which operating modes are supported: teleoperation, recorded routes, waypoint navigation, map-based SLAM, dynamic obstacle avoidance and autonomous recovery. These are different functions. A successful route-following demonstration does not show robust localization after a layout change or safe behavior after a sensor fault.
Run a test matrix with genuine obstacles, people, narrow passages, reflective surfaces, lost localization, blocked routes and interrupted connectivity. Define allowed intervention, safe-stop behavior and time to resume. Record software version, logs and environmental conditions.
03 / Treat docking and endurance as reliability topics
Autonomous docking is a process, not a single successful video clip. Test repeated docking attempts under realistic floor, lighting and battery conditions; measure success rate, retries and intervention time. A claim such as “fully autonomous charging” must include how the robot detects, reaches and recovers from a failed docking event.
Compute the energy budget for a full shift, including motion, waiting, sensing, networking, payload operation and reserve. Review battery aging, charging temperature limitations, queueing and what happens when a dock is unavailable.
04 / Verify protection, safety and conformity within scope
An IP rating is meaningful only for the tested configuration and relevant openings, connectors, sensors and charging contacts. Clarify whether external modules or field modifications change that protection. Request the actual test scope and any operating limitations.
Do not treat a list of directives or standards as a universal checklist. Applicable EU requirements depend on the product, radio functions, voltage, intended use and how the complete machinery system is placed on the market. Review the relevant declarations, technical documentation, risk assessment and test evidence with competent specialists.
05 / Challenge software, support and lifecycle assumptions
Verify what the SDK actually exposes: mission control, telemetry, fault codes, map management, event subscriptions, safety-related limits and integration recovery. Ask who owns the data, where it is stored, which traffic leaves the site and whether a validated on-premises deployment is possible.
Build a maintenance and spares picture before the purchase order: batteries, joints, LiDAR, tyres, docks, firmware support, diagnostic tools, replacement lead times and warranty exclusions. A competitively priced robot with critical parts unavailable locally can become an expensive operational dependency.
Technical due diligence does not slow innovation. It turns an impressive technology demonstration into a defendable industrial investment.