Preparing the published article.
Wheeled AMR or Legged Robot: How to Compare Before You Buy
Measure route segments, cycle times and intervention rates, then run a side-by-side trial before choosing wheels or legs.
Most site teams start by asking which platform is better. The productive question is narrower: for each segment of the work you actually need done, is mobility the binding constraint, or is something else?
1. Break the task into segments before comparing platforms
Write the job as a numbered route, not a job title. For a typical internal goods or inspection round that means: pickup point, travel A, threshold or door, travel B, turn or backing move, destination, dwell time, return. For each segment record the floor type (epoxy, bare concrete, grating, carpet tile, expansion joint), the largest lip or threshold height in millimetres, the narrowest door or aisle width, the ramp or slope, whether the path is shared with pedestrians or forklifts, and whether any part of the route leaves the building. Measure these on site with a laser distance meter and a straightedge; do not take them from a drawing. Then mark each segment as trivial, marginal or blocking for (a) a wheeled AMR and (b) a legged platform. The comparison stops being a debate about technology and becomes a list of two or three segments that actually decide the choice.
Record each segment in one row of a table with columns: segment id, floor type, lip height, aisle width, shared traffic, and verdict for each platform. Anything you cannot measure, mark as unknown and put it on the site-survey list, because an unmeasured threshold is the most common reason a trial fails after purchase.
2. Compare throughput, intervention rate and energy logistics
Cycle time is the honest comparator, expressed as a median and a worst case over a full shift, not a best-run figure. Ask what the platform's steady-state speed is on your floor type, how long a door or lift interaction takes, and what happens when a person blocks the path. Then ask the intervention question: what does the platform do when it is stuck or has lost localisation, how long does a manual recovery take, and who performs it. For legged platforms, ask specifically how many falls or near-falls per hundred operating hours the integrator expects, and what recovery looks like after one. For wheeled platforms, ask for the floor condition limits in writing and what the warranty position is if the floor degrades. Energy logistics decide the round: measure the distance from the operating area to a usable charge point, confirm who owns that space, and compare opportunity charging against battery swap. A platform that needs two hours of downtime in a ten-hour shift is a different proposition from one that needs twenty minutes, regardless of which one walks stairs.
Put the two platforms side by side in one table: median cycle time, worst cycle time, interventions per hundred cycles, recovery minutes per intervention, and charge logistics. Any cell the supplier will not commit to in writing should be flagged as an unverified claim, not treated as a specification.
3. Run a side-by-side trial with a log you own
Run both platforms on the same route, same payload, same shift pattern, for long enough to include a busy day and a slow one. Define one trial week with identical tasks and a written stop rule: if an intervention rate exceeds the threshold you set, the trial pauses and the event is reviewed rather than absorbed. Keep the log in your own spreadsheet, not the vendor dashboard, with fields for timestamp, segment id, task outcome, intervention cause code, recovery duration, and the name of the person who intervened. Photograph every floor defect or threshold that caused a stop, with a ruler in frame. Before signing, ask for the maintenance items your own technicians will have to perform, the spare parts that should sit on site, and the training hours required per operator. The decision rule should be written before the trial begins: state which metric, at what level, would make you choose the legged option, and which would keep you on wheels.
Photo credits: the images in this article are licensed reference photographs of the relevant robot class, credited with author and licence in each caption; they are not photographs of the specific equipment discussed. Licence and author details are recorded with each image.