humanoid4rent
登录获取报价

Service robots in European airports, shops and hotels: what the pilots reported

Field studies at Schiphol, Finnish shops, Kraków cafés and a hotel show curious but cautious visitors, slow conversation and heavy staff support. Cost data is almost never published.

作者: humanoid4rent 编辑团队 · 发布于 2026年11月12日

本页面尚未翻译
您看到的是英文原文,因为目前还没有中文版本。没有任何内容被省略:这是完整的页面,以其撰写时使用的语言呈现。

Published results from European service robot pilots are thinner than the headlines suggest. Field studies at Amsterdam Schiphol, a Finnish shopping centre and grocery store, two Kraków cafés and a Dutch-led hotel trial describe visitors who are curious but cautious. Slow conversation is the most common complaint, and staff support behind the scenes is substantial. Almost none of them publish operating cost figures.

This article collects what has been documented in peer-reviewed papers and official EU project records, including pilots that have ended. It focuses on two questions organisers and venue managers ask most often: how did real visitors react, and what did it take to run the robot?

Why the evidence base is small

Many robot appearances in shops, terminals and hotel lobbies are announced but never evaluated in public. The deployments with published outcomes are mostly research projects, where universities measured reactions and wrote up their findings. That has two consequences. Sample sizes are often small, and the robots were frequently supported by researchers or controlled remotely.

The results below should therefore be read as indications, not as benchmarks. Where a study did not report something, such as cost, this article says so rather than filling the gap.

Airport: a guide robot at Schiphol

The EU-funded SPENCER project ran from April 2013 to March 2016 and received about €3.18 million from the EU, according to the CORDIS project record. Its goal was a robot that could guide transfer passengers at Amsterdam Schiphol Airport, with KLM as the end-user partner.

A field evaluation by the University of Twente, published at the HRI 2017 conference, gives the clearest picture of how passengers responded. In Joosse and Evers' paper, 16 passengers in nine groups followed the robot autonomously along a route of roughly 430 metres, first through a busy shopping area and then along quieter corridors.

The reported findings were mixed:

  • Of 13 participants asked, 12 said a service like this would improve their satisfaction; one was undecided.
  • Speed divided opinion: three people wanted it faster, while five thought it was too fast for families.
  • Passengers liked that the robot stopped when people came close, but in crowds it stopped too often and re-planned slowly, so they saw it as less suited to congested areas and passengers in a hurry.
  • Many said the robot was useful, but mainly for other people, such as older travellers and families.

The authors also noted that the trial was particularly resource-intensive, because extra staff were needed at the airport for safety and security. They acknowledged that participants may have answered positively to please the researchers.

Shopping centre: Pepper at Ideapark in Finland

The MuMMER project put a Pepper humanoid into the Ideapark shopping centre in Lempäälä, Finland, to chat with visitors, play games and give directions. The CORDIS results summary reports a final autonomous deployment on 49 days over 14 weeks, from September 2019 to January 2020, after shorter visits in earlier years. The EU contribution was about €4.3 million.

The summary refers to acceptance studies with several hundred participants in earlier phases, but it does not publish visitor counts or satisfaction scores for the final deployment. For anyone looking for a headline figure, there is none in the official record. The project's main published result is the technical system itself, combining speech, vision, social signal processing and navigation that takes people into account.

If you are considering a similar attraction, our guide to robots in shopping centres and retail covers the practical side.

Grocery store: two robots in a Finnish supermarket

Researchers at Arcada University of Applied Sciences in Finland tested two social robots in a grocery store and presented the results at the ICEIS 2025 conference. According to the paper's abstract, workshops and trials produced seven possible roles, from store guide and sales promoter to entertainer and store chef.

The team surveyed 39 customers. Most viewed the robots positively, and the robots carried out their tasks reasonably well. However, the authors stress that a notable share of shoppers found social robots unpleasant in a store. The sample is small, and the abstract does not name the robot models or the retailer.

Cafés: a remotely controlled NAO in Kraków

A 2023 field study by researchers from AGH University, Jagiellonian University and Université Paris Cité placed a small NAO humanoid next to the counter in two Kraków cafés. As the paper describes, a hidden operator typed the robot's replies, while customers were led to believe it was autonomous.

When the robot only made small talk near the queue, most visitors just took photos or greeted it. In a two-hour session the researchers estimated about 50 visitors, of whom five held a conversation. Engagement improved once NAO was given a clear job: taking orders and passing them to the barista.

The weak points were specific:

  • Replies were slow, and five of nine interviewees in the final sessions complained about this. Some customers gave up and ordered from the human barista instead.
  • Some people found the robot's childlike voice unsettling, and some doubted it would understand them.
  • Café noise made it hard for the operator to hear customers, because the setup had no microphone for the operator.

The authors summarised Polish customers' attitude as neutral and uncertain rather than hostile. For more on how remote control changes what audiences see, read our guide to teleoperated versus autonomous robots.

Hotel: robot and human side by side

A field experiment led by Hotelschool The Hague, published in Strategic Change in January 2025, compared a human and a robot at an information point in a working hotel. According to the article's abstract, 200 guests took part, and both agents were available at the same time.

The researchers found no meaningful difference between human and robot in guest satisfaction or intention to return. They suggest robots could support staff on routine information tasks. The paper presents this as a possible cost benefit rather than a measured one, and the authors note that the study covered only one touchpoint. We were only able to read the abstract, which does not name the robot model.

Our guide to robots in hotels and hospitality explains how short event bookings differ from pilots like this.

Beyond the novelty: six weeks with a robot barista

Most of the trials above lasted hours or days. A rare longer study from Heriot-Watt University in Edinburgh ran a Furhat robot head, connected to a coffee machine, in a departmental common room for six weeks. Lim and colleagues report that 53 people signed up and 186 interactions were recorded. Users rated their satisfaction at 7.25 out of 10 on average, and an interaction typically took about two and a half minutes.

The researchers found that trust and people's prior attitudes towards robots were linked to how useful and likeable they found the barista. Their first practical lesson was blunt: people get interrupted, leave mid-conversation and do not wait for a robot to finish its script. Only 22 participants completed the exit questionnaire, which shows how quickly participation can thin out.

Operating costs: what is published

None of the studies above publishes running costs, staffing hours or revenue effects in figures. What they do document points one way:

  • The Schiphol trial needed extra airport staff for safety and security.
  • The Kraków robot needed a hidden human operator for every conversation.
  • The Edinburgh barista needed a registration system, loyalty cards and questionnaires to identify users.

The hotel study's cost argument is an inference from equal satisfaction scores, not a measured saving. Anyone presented with a return-on-investment claim from a pilot should ask whether staff time for supervision, setup and remote operation was counted. Our guide on measuring the return of a robot at an event sets out what to track.

The platform behind many pilots changed hands

Several of these trials used Pepper or NAO, robots originally developed by Aldebaran in Paris. Heise reported in July 2025 that Aldebaran had filed for insolvency in February 2025 and that the Chinese company Maxvision Technology had acquired its technology and intellectual property. Maxvision said it planned a French subsidiary for development and customer service. The price was not disclosed.

For venues and suppliers that already own these robots, the practical question is long-term support for spare parts and software. Our Pepper catalogue page lists the robot's published capabilities.

What this means for event organisers

  • Give the robot a clear job. In Kraków and at Schiphol, people engaged more when the robot did something specific, such as taking an order or guiding to a gate, than when it simply chatted.
  • Plan for slow replies and noise. Response delay was the most common complaint, and café noise defeated the setup. Our guide to speech, languages and noisy venues covers microphones and scripting.
  • Budget for people, not only the robot. Every documented pilot relied on operators, researchers or extra staff. Ask suppliers exactly who will be on site.
  • Expect a mixed audience. Even favourable studies found a share of visitors who were uneasy, so leave room for people to keep their distance.
  • Treat pilot claims with care. Small samples, a single touchpoint and no cost data are the norm, so ask how any figure you are shown was measured.

Frequently asked questions

Did the Schiphol guide robot trial show passengers liked robot guidance?

Mostly, but the sample was small. In the University of Twente field study, 16 passengers in nine groups followed the SPENCER robot about 430 metres. Twelve of the 13 asked said such a service would improve satisfaction. They also said it stopped too often in crowds and re-planned slowly, and they saw it as more useful for families and older travellers than for themselves. Extra airport staff were needed for safety and security.

Do European robot pilots publish operating cost data?

Very rarely. None of the field studies reviewed here, covering an airport, a shopping centre, a grocery store, cafés, a hotel and a robot barista, published running costs or staff hours as figures. They do show that robots relied on operators, researchers or additional staff. A 2025 hotel study suggests possible savings because guest satisfaction was equal, but that is an inference rather than a measured cost.

Why did customers in the Kraków café study lose interest in the robot?

Mainly because replies were slow. The NAO robot was controlled by a hidden operator who typed its answers, and five of nine interviewees in the final sessions complained about delays. Some customers gave up and ordered from the barista. Café noise also made it hard for the operator to hear. Engagement improved when the robot had a clear task, such as taking orders, rather than only making small talk.

Are hotel guests less satisfied when a robot serves them instead of a person?

One field experiment suggests not, at least for simple tasks. A study led by Hotelschool The Hague, published in January 2025, compared a human and a robot at a hotel information point with 200 guests. It found no meaningful difference in satisfaction or intention to return. The authors note that it covered only one touchpoint, so the result may not apply to check-in, complaints or other complex interactions.

How this article was made: written by the humanoid4rent editorial team with AI assistance; every factual statement was checked against the sources listed below on 2026-11-12. Spotted an error? Write to info@humanoid4rent.com.

资料来源

本文中的事实已在撰写当天对照以下页面进行核实。

  1. Social situation-aware perception and action for cognitive robots (SPENCER) – CORDIS, European Commission
  2. A Guide Robot at the Airport: First Impressions – Joosse and Evers, HRI 2017 Companion (University of Twente)
  3. MultiModal Mall Entertainment Robot (MuMMER), reporting – CORDIS, European Commission
  4. Customer-Facing Social Robots in the Grocery Store: Experiences from a Field Trial – ICEIS 2025, SciTePress
  5. A field study on Polish customers' attitude towards a service robot in a cafe – Kiraga, Samsel and Indurkhya, arXiv
  6. Human Versus Robot: Comparing Service Agents in Hospitality Settings—Insights From a Field Study – Strategic Change, Wiley
  7. Feeding the Coffee Habit: A Longitudinal Study of a Robo-Barista – Lim, Robb, Wilson and Hastie, arXiv (RO-MAN 2023)
  8. Pepper und Nao gerettet: Maxvision kauft insolvente Robotik-Firma Aldebaran – heise online

博客更多内容

您的下一场活动想要一台机器人吗?

只需一次性说明日期、地点以及机器人需要完成的任务。供应商会给您回复报价,方便您逐项对比。

发送需求

您从事机器人行业吗?欢迎为我们的博客撰稿。 →