When wheels won’t do: Humanoid robots for human-centric spaces
The warehouse floor is a landscape of contradictions. It is at once meticulously organized and chaotically uneven, a place where the predictable geometry of a conveyor belt meets the unpredictable reality of a pallet jack skid, a dropped box, or a slightly warped floor tile. For decades, the default answer to material handling automation has been the wheel. It is efficient, it is proven, and it is utterly defeated by a modest step or a stretch of cracked concrete.
This is the central thesis behind a new wave of robotic design that is less concerned with speed on a flat surface and more concerned with the simple, human act of walking. The argument, put forward by developers like Agility Robotics, is that our built environment—the warehouses, the factories, the construction sites—was designed for bipedal locomotion. We did not build these spaces for robots; we built them for people. Therefore, the most effective robot for these spaces might be the one that moves the way we do.
This guide examines the practical implications of this shift, looking at what to consider when evaluating humanoid platforms, how to approach implementation, and where the common pitfalls lie. We are moving beyond the laboratory and into the realm of operational logistics, where the question is no longer "can it walk?" but "can it work?"
What to look for
When evaluating a humanoid robot for a human-centric environment, the first and most obvious criterion is physical form factor. The industry is not converging on a single size, but rather on a range that reflects the diversity of human tasks. Consider the example of Agility Robotics' Digit. This is a machine designed to be roughly the size of an adult human—standing at 1.75 meters tall and weighing in at 65 kilograms. This is not an arbitrary choice. The dimensions allow the robot to navigate doorways, reach standard shelving, and operate within the same spatial envelope as a human worker.
The weight is a critical factor. At 65 kilograms, Digit is heavy enough to be stable but light enough to be manageable. The payload capacity, in this case 16 kilograms, tells you the intended scope of work. This is not a heavy-lifting machine; it is a handling machine. It is designed to move totes, components, and other items that a human would typically carry. When you look at a humanoid, you must match the payload to the task. If your operation requires moving 30-kilogram loads, a 16-kilogram capacity robot is not the answer. The source material is explicit about this specific capacity, but it does not disclose whether higher-capacity variants are planned. That is a detail that remains undisclosed and should be a direct question to any vendor.
The second thing to look for is the specific use case justification. A humanoid robot is not a general-purpose replacement for all automation. The developers at Agility Robotics are clear on this point. They reference the "empty tote recycling" task as a prime example. In many warehouses, this task is already automated using a conveyor system. The conveyor is a "very traditional automation solution," and it works well—but only in specific spaces. A conveyor has a fixed footprint. It requires a certain layout, a certain flow, a certain clearance. When your warehouse layout deviates from that ideal, the conveyor becomes a liability rather than an asset.
This is where the humanoid robot earns its place. It is not meant to replace the conveyor where the conveyor works. It is meant to operate in the spaces where a conveyor is not feasible. Look for the "functional need to exist." If a task is already handled efficiently by a fixed automation solution, a humanoid is likely a downgrade in cost and complexity. If, however, you have a task that requires mobility across uneven terrain, or the ability to navigate around obstacles that would stop a wheeled robot, then you have a candidate.
The third aspect to examine is the terrain capability. The advantage of bipedal locomotion is not that it is faster than wheels—it is not. The advantage is that it is more adaptable. The source material quotes a representative from Agility Robotics, Vetterick, who notes that our buildings are easy for people to navigate, but even small obstructions become barriers to a wheeled robot. A walking robot, however, can step over or around these obstructions. When evaluating a robot, do not just look at the spec sheet for top speed. Look at the step height, the ability to handle gradients, and the stability on uneven floors. The source material explicitly mentions "uneven floors" as a key differentiator, but it does not provide specific technical specifications for the maximum step height or incline angle. This is a critical gap in the public data, and you should request this information directly from the manufacturer.
Finally, look at the timeline for availability. The humanoid market is not monolithic. While Digit is a current platform, others are on the horizon. The source material introduces Zyrex, a construction robot from RIC Robotics, which is slated for a 2026 introduction. This is a 20-foot tall machine—a stark contrast to the human-scale Digit. This highlights that "humanoid" is a broad category. It can refer to a robot that mimics human size, or it can refer to a robot that mimics human form in a scaled-up, industrial context. When planning your automation roadmap, you must be aware of this spectrum. A 20-foot tall robot is not going to work in a standard warehouse, but it might be relevant for large-scale construction projects. The source material does not provide further details on Zyrex's payload, power source, or specific construction tasks, so those remain open questions for the vendor.
Practical steps
The first practical step is to audit your facility for "conveyor-incompatible" tasks. Walk your warehouse, your factory floor, or your construction site and identify the operations that are currently done manually because automation has failed to fit. Look for the corners, the odd-shaped rooms, the areas with uneven flooring. These are the zones where a wheeled robot has been tried and failed, or where a conveyor was never even considered due to space constraints. List these tasks. Quantify the time spent on them, the labor cost, and the error rate. This is your baseline.
The second step is to match the robot to the task, not the other way around. Do not buy a humanoid and then try to find a job for it. Start with the job. If you have an empty tote recycling task that is currently manual because a conveyor footprint is not available, that is a perfect candidate for a robot like Digit. The robot's 16-kilogram payload is suitable for totes. Its 1.75-meter height allows it to reach standard workstations. Its walking capability allows it to navigate the uneven floor that blocked the wheeled alternative. If your task involves heavy lifting, a humanoid is likely the wrong tool. The source material does not suggest that these robots are suitable for all tasks, and you should not assume they are.
The third step is to pilot in a constrained environment. Do not roll out a fleet of humanoids across your entire operation on day one. Select a single, well-defined task in a specific area. Set clear metrics for success: throughput, error rate, uptime. Run the pilot for a sufficient period to gather data. The source material does not specify a typical pilot duration, so you will need to negotiate this with the vendor. The key is to isolate the variable. You want to know if the robot can perform the task in your environment, with your floor, your lighting, and your workflows.
The fourth step is to redesign the workflow around the robot's strengths. A humanoid robot is not a drop-in replacement for a human. It does not have the same dexterity, the same cognitive flexibility, or the same ability to adapt to novel situations. You must adapt the workflow to suit the robot. For example, if the robot is tasked with moving empty totes, ensure that the totes are placed in a consistent, predictable location. Ensure that the drop-off point is clear of obstructions. The robot works best in a semi-structured environment. The source material indicates that these robots are designed to "fit into spaces and workflows designed for humans," but this does not mean they can handle chaos. It means they can handle the *layout* of a human space, not the *disorder* of a human workspace.
The fifth step is to plan for the physical interface. How will the robot charge? How will it be serviced? Where will it be stored when not in use? These are logistical questions that are often overlooked. A 65-kilogram robot needs a stable charging station. A 20-foot tall robot needs a completely different infrastructure. The source material does not provide details on charging times or battery life for either Digit or Zyrex. This is a critical unknown. You must ask the vendor for these specifications before committing to a purchase. You must also plan for maintenance. Who will service the robot? Will it be your in-house team, or will it require vendor technicians? The source material does not disclose the service model, so this is another point of negotiation.
The sixth step is to consider the software integration. A robot is only as useful as the data it receives. How will the robot know where to go? How will it know which totes to pick up? The source material does not detail the software stack for either robot. You should assume that integration with your Warehouse Management System (WMS) or Enterprise Resource Planning (ERP) system will be a project in itself. Plan for this in your budget and timeline.
Common mistakes to avoid
The most common mistake is assuming that a humanoid is a universal solution. The source material is clear that even the developers do not believe this. They explicitly state that they "wouldn’t want a robot in a place where a conveyor works." If you have a task that is currently automated by a conveyor, and that conveyor is working well, do not replace it with a humanoid. You will be introducing complexity, cost, and potential downtime for no benefit. The humanoid is for the spaces where the conveyor cannot go.
The second mistake is ignoring the terrain. The primary selling point of a bipedal robot is its ability to handle uneven floors. If your facility has perfectly smooth, flat floors, a wheeled robot is almost certainly a cheaper, faster, and more reliable option. The source material explicitly contrasts the two, noting that a wheeled robot might struggle with "small obstructions" that a walking robot can handle. If you do not have those obstructions, you do not have a use case for a walking robot. Do not buy a humanoid just because it is technologically impressive. Buy it because you have a specific problem that only a walking robot can solve.
The third mistake is underestimating the payload limitations. A 16-kilogram payload is not a heavy-duty specification. It is a light-to-medium duty specification. If you try to overload the robot, you will damage it or cause it to fail. The source material provides this specific number for Digit, and you must respect it. Similarly, the 20-foot tall Zyrex is described as a "construction robot," but the source material does not provide its payload capacity. Do not assume that bigger means stronger. You must get the specific payload rating from the vendor before you plan any tasks for it.
The fourth mistake is failing to plan for the "conveyor footprint" problem. The source material notes that a conveyor has a "specific footprint" and is "conducive to certain types of spaces." When you remove a conveyor, you free up that footprint. You must have a plan for what goes into that space. Is it a new workstation? A storage area? A path for the robot? If you do not plan for the freed space, you may find that the robot's path is blocked by the very equipment you removed to make room for it.
The fifth mistake is expecting human-level autonomy. These robots are not androids. They are not capable of understanding complex, ambiguous instructions. They are designed for specific, repetitive tasks. The source material describes Digit as a "multipurpose" robot, but "multipurpose" in this context means it can be reconfigured for different tote-handling tasks, not that it can perform any task a human can. You must set realistic expectations with your workforce and your management. The robot is a tool, not a replacement for human judgment.
The sixth mistake is ignoring the timeline. The source material notes that Zyrex is coming in 2026. That is a future date. If you are planning a construction project that requires a 20-foot tall robot, you cannot buy one today. You must plan your project timeline around the robot's availability. Similarly, if you are looking at Digit, you must check the current lead times for delivery. The source material does not provide these lead times, so you must contact the vendor directly. Do not assume that you can order a robot and have it on your floor next week.
The seventh mistake is neglecting the total cost of ownership. The purchase price of the robot is only the beginning. You must factor in maintenance, software licensing, training, and potential downtime. The source material does not provide pricing for either robot, so you must obtain quotes from the vendors. You must also consider the cost of the infrastructure changes required to support the robot. A charging station, a maintenance bay, and potentially reinforced flooring are all costs that must be budgeted.
The eighth mistake is failing to engage the workforce. Introducing a humanoid robot into a warehouse will raise questions and concerns among your employees. They may worry about job security. They may be skeptical about the robot's capabilities. You must communicate clearly about the robot's role. The source material suggests that the robot is intended for tasks that are not currently automated, such as empty tote recycling in spaces where a conveyor is not feasible. This is a task that is likely being done manually today. You must explain to your employees that the robot is taking over a task that is physically demanding and repetitive, and that it is not replacing them but rather augmenting their capabilities. The source material does not provide guidance on labor relations, but it is a practical consideration that cannot be ignored.
The ninth mistake is assuming that all humanoids are created equal. The source material presents two very different robots: Digit, a 1.75-meter, 65-kilogram human-scale robot, and Zyrex, a 20-foot tall construction machine. These are not competitors in the same market. They are solutions for entirely different problems. You must be clear about which category of robot you need. If you need to navigate a standard warehouse, a 20-foot tall robot is useless. If you are working on a high-rise construction project, a 1.75-meter robot is not going to be able to reach the upper floors. Match the robot's scale to your operational scale.
The tenth mistake is ignoring the "functional need to exist." The source material uses this phrase to describe the conditions under which a robot is necessary. A robot has a "functional need to exist" when there is no other automation solution that can perform the task. If a conveyor can do it, use a conveyor. If a wheeled robot can do it, use a wheeled robot. Only when those options fail should you consider a humanoid. This is the core principle of the entire article. Do not let the novelty of the technology cloud your judgment. The goal is not to have a humanoid robot; the goal is to solve a logistics problem. The humanoid is simply one possible tool.
Sources
When wheels won’t do: Humanoid robots for human-centric spaces
Published by Vigla Media OÜ (Estonia).