Limits of robots
What can a humanoid robot not do yet?
· 5 min read
In videos a humanoid robot looks almost human today, but reality has its limits. If you are considering a rental, it is important to know what the robot cannot do, so that you plan according to facts and not according to marketing edits.
Put simply: what not to expect from a robot
Think of the robot as a very skilful but literal assistant. It handles exactly what it was prepared and trained for, but outside that frame it flounders. It cannot understand an unfamiliar situation on its own the way a person does in a few seconds, and it has no common sense to fill in the gaps.
In practice this means one simple thing. A humanoid robot is excellent for learnt, repeatable and pre-prepared scenarios. Improvisation, solving completely unexpected problems and full independence without the operator's supervision are still beyond what today's technology can reliably handle.
Most of the impressive videos circulating online were made in a controlled environment, after many attempts and often with a human operator in the background. It is not a scam, it is the industry standard. What matters is knowing that there is a difference between an edited video and a live deployment at your event, and it needs to be taken into account.
Fine motor skills and grip: the biggest weakness
The human hand can adapt to thousands of shapes, surfaces and weights without us thinking about it. Robotic hands do not yet have this feel. They lack sufficient touch sensitivity, fine control of force and the ability to grip a variety of objects securely.
That is why a robot reliably picks up a predefined cup or leaflet but struggles with a soft, slippery or irregular object. Handing over fragile glass, unwrapping foil or working with things that change shape should be treated as a risk area, not as a given.
There is also a difference in feedback. A person knows immediately by touch whether an object is slipping out of their hand and adjusts the force. A robot can only handle this continuous correlation to a limited extent so far, so tasks such as gentle pouring, folding or sorting small items are among those where realistic expectations are needed and everything should be tested in advance.
- Reliable: learnt grips of prepared objects with a fixed shape
- Risky: soft, slippery, fragile or irregular objects
- Not yet: rapid adaptation of the grip to a completely unknown object
How it really works: why stability is expensive
Walking on two legs does not come free. Every fraction of a second the robot has to correct its balance, plan its movement and react to the surface under its feet. This constant work costs energy and computing power, which directly affects how long and in what environment the robot can keep working.
On a level, clean and predictable floor the movement is smooth and impressive. Uneven terrain, stairs in a crowd, wet tiles, deep-pile carpets or a crush of people are demanding, however. In such an environment the risk of a fall grows and the robot needs more open space as well as the operator's assistance.
This is also why wheeled robots often win in the industry where all that is needed is moving across a smooth floor. The humanoid shape makes sense mainly in spaces built for people and where the human silhouette itself is part of the experience. For an event that works in its favour, the space just has to be adapted to the robot.
Battery life and independence: the hard numbers
The battery is the quiet limit of the whole industry. Humanoid robots of the Unitree G1 class, which you know from our fleet as MONTARI and NOVA, run for roughly two hours per charge in active deployment, and even less with intensive arm work. Recharging then takes around an hour and a half.
The second limit is independence. Despite the label of a general-purpose robot, today's models handle a narrow set of tasks and full autonomy without a human is still a matter of the coming years. That is why at events we work with an operator who controls the robot, supervises it and deals with unexpected situations as they arise.
Speech, context and decision-making
The robot can answer, gesture and carry out a prepared interaction, but it does not understand the world as deeply as a person. It cannot reliably assess intent, the mood of the crowd or hidden context, and its reactions are best where the scenario has been prepared and tested in advance.
Complex decision-making in uncertain conditions, independent recovery from an error in the middle of a task or reliable work in a completely unstructured environment are areas that research is still working on. For the client this leads to a clear conclusion. The better the scenario is prepared, the better the result.
A practical example. A robot as a host at a stand that welcomes visitors, says prepared sentences, poses for photos and responds to simple commands works brilliantly. A robot that is supposed to hold a free, unguided discussion with a random person about anything runs into its limits and seems less convincing. The difference lies in the degree of preparation.
What a company should take from this
The limits are not a reason to say no, they are a guide to a good brief. A robot has a huge effect where the added value is the very presence of impressive technology: product launches, trade fairs, corporate events, filming and attracting attention. There it delivers the wow effect and topics for conversation.
On the other hand, do not expect a full replacement for staff, independent operation without supervision or flawless handling of random objects. If you want to go through the options and formats calmly in advance, take a look at the rental types and choose the scenario that fits your goal and space.
Our advice is sober. Define a clear scenario, give the robot a prepared environment, plan for an operator and for charging breaks. With this approach the robot reliably delivers what is really needed from it, and we do not promise you anything the technology cannot yet carry.
And some good news to finish. Development is moving forward quickly, fine motor skills and independence are improving year on year, and the models that are coming will raise the bar. That is exactly why we watch the market and keep adding to the fleet. Until then, we would rather say straight out where the limits are, so that at your event you do not experience disappointment but the effect you expected.
The limits summarised on one page
If you take one paragraph away from this article, let it be this one. A humanoid robot in 2026 is a top-class tool for attention, presentation and experience in a prepared environment, but it is not an independent workforce. Its weaknesses remain fine grip, battery life, movement over demanding terrain and independent decision-making in unfamiliar situations.
- Strengths: presence, attention, prepared interaction, wow effect
- Weaknesses: fine manipulation, endurance, terrain, full independence
- Solution: a prepared scenario, an operator, charging breaks
