Universities Face Potential Research Crisis as Popular Robotics Company Struggles

Universities face getting stuck with thousands of obsolete robots - here's how to avoid a research calamity

For more than a decade, the French robotics company Aldebaran has built some of the most popular robots used in academic research. Go to most university robotics departments and you'll find either Pepper, the iconic three-wheeled humanoid robot, or its smaller two-legged sibling, Nao. These fast became the robots of choice for many academics for all research into the capabilities and potential of social robots. They are quick to set up and easy to use out of the box, without the need for any programming skills or engineering knowledge. With base prices at around £17,000 for Pepper and £8,000 for Nao - typically plus a few thousand pounds more for extras, online training sessions, service plans, warranties and so on - the robots could be purchased via university research grants. With Pepper robots also appearing in customer service jobs, for example in HSBC banks across the US, buyers were attracted by the lure of long-term educational and financial benefits from a state-of-the-art tech supplier. Aldebaran says it has sold approximately 37,000 machines worldwide (20,000 Naos and 17,000 Peppers). Get your news from actual experts, straight to your inbox. Sign up to our daily newsletter to receive all The Conversation UK's latest coverage of news and research, from politics and business to the arts and sciences. However, the company stopped developing Pepper robots in 2021, having struggled to sell as many as it had hoped, and was offloaded by long-time Japanese owner Softbank. In February of this year, Alderbaran filed for bankruptcy and restructured amid ongoing financial difficulties. Currently looking for a buyer, it has halved its staff numbers, though it is still making Nao (and a serving assistant on wheels called Plato). The uncertainty around the company's future has stoked fears that it will become impossible to get its robots repaired in future, and that Aldebaran could stop supporting the AI cloud network that the machines need to access to be able to function. What does this mean for the future of robotics research in universities? Besides fears about Aldebaran's future, there have long been issues with Pepper and Nao's durability. They both have rigid, fragile plastic shells, and the machines sometimes overheat. This means they have to be left to cool down after 20-30 minutes, which has often interfered with experiments and data-gathering - as documented in this 2022 study of Nao. A spokesperson for Aldebaran agreed that motors can overheat, depending on their use and environment. They said the next generation of Nao, currently in development, has taken this into account in its design. For repairs, the only option is Aldebaran or an authorised reseller, or you risk voiding your warranty. This typically involves shipping overseas, which can be slow and costly - more so if the replacement parts are out of stock. One of us (Emilia) encountered this during the COVID pandemic. Nao's batteries need to be used regularly to keep functioning, which led the university's machine to fail because it was inaccessible during lockdowns. Aldebaran couldn't supply replacement batteries quickly, which halted research projects at the university for many months and meant that important submission deadlines were missed. Meanwhile, software upgrades for Pepper stopped when the company halted development in 2021 (sales stopped in 2024). This robot's limited processing capabilities make it troublesome to run the large language models (LLMs) that power interfaces like ChatGPT (although these can be run in conjunction with a computer with modifications). Nao does have an AI edition that can handle LLMs, though this too requires external modifications. Nao's upgrades also seem to have been limited, which in our experience appears to have made them more error-prone too. Both robots are already considerably less useful for research purposes in our opinion. Finally, Nao and Pepper were not built with adaptability in mind. Unlike more recent machines like the 3D-printed InMoov, made by French designer Gael Langevin, there's no way of customising their components or appearance. Their fixed expressions, gestures and plastic body make them difficult to adapt to different user needs or applications, such as helping at home or in healthcare. This again reduces their usefulness from a research point of view. Addressing these concerns, the Aldebaran spokesperson said: Spare parts availability on Nao is very good, [barring] the normal supply chain issues, and these were exacerbated during COVID like the rest of the commercial world. Pepper is more limited as it has not been in production for some time, but we are generally able to solve any customer issues. Nao is still very active as a product, with production continuing along with software upgrades. We recently launched Nao Activities, a major software upgrade that provides generative AI capabilities for Nao. The spokesperson added that are were no plans to switch off AI cloud support for Nao or Pepper, and that the robots are not difficult to use in robotics research, "testament of which is the thousands of units being used in that environment". What can be done? If Pepper and Nao do become unusable for research, universities will have to either scrap them or try to redevelop them with custom parts and components. It's possible they could be hacked and gutted, replacement parts could be 3D-printed, new microprocessors installed and the software made local and open source, which may be enough to get the robots back up and working again. However, it probably makes sense for researchers to look forwards instead. But towards what? At a time when university finances are very tight, there may be a reluctance to buy new machines with potentially limited shelf lives. Robots from alternative providers such as Futhat and Unitree are supported by similar cloud-based AI systems. Some institutions may consider reallocating vital funding to other departments, with a significant impact across robotics research and education. Universities are at the heart of robotics research, upholding high ethical standards and rigorously testing machines without the conflicts of interest that manufacturers can have. Universities can also bring together diverse disciplines like computer science, engineering and cognitive science, fostering collaboration that encourages innovation. With the UK number one globally for research quality in this field, these are the training grounds for the next generation of roboticists at a time when there is a growing skills shortage. A different way forward would be for universities to start building and programming robots from scratch. For the cost of a new research robot, say £15,000, you could buy several high-spec 3D printers, hardware and components. This wouldn't be about building entire humanoid robots but prototypes of key aspects such as facial expressiveness or skin, human gestures or emotions. This would allow students to gain important hands-on engineering and programming skills, while conducting novel research exploring current gaps in the field. It would make personalising them easier and repairing them quicker and cheaper, if you could 3D-print parts or use parts that could be easily replaced off-the-shelf. If universities are to remain relevant in this rapidly evolving field, it's vital that they learn from their difficulties with Pepper and Nao. At a time when robots are starting to be perceived as reliable and cost-effective support for people, this is a cautionary tale for all.

Universities face getting stuck with thousands of obsolete robots. Here's how to avoid a research calamity

For more than a decade, the French robotics company Aldebaran has built some of the most popular robots used in academic research. Go to most university robotics departments and you'll find either Pepper, the iconic three-wheeled humanoid robot, or its smaller two-legged sibling, Nao. These fast became the robots of choice for many academics for all research into the capabilities and potential of social robots. They are quick to set up and easy to use out of the box, without the need for any programming skills or engineering knowledge. With base prices at around £17,000 for Pepper and £8,000 for Nao -- typically plus a few thousand pounds more for extras, online training sessions, service plans, warranties and so on -- the robots could be purchased via university research grants. With Pepper robots also appearing in customer service jobs, for example in HSBC banks across the US, buyers were attracted by the lure of long-term educational and financial benefits from a state-of-the-art tech supplier. Aldebaran says it has sold approximately 37,000 machines worldwide (20,000 Naos and 17,000 Peppers). However, the company stopped developing Pepper robots in 2021, having struggled to sell as many as it had hoped, and was offloaded by long-time Japanese owner Softbank. In February of this year, Alderbaran filed for bankruptcy and restructured amid ongoing financial difficulties. Currently looking for a buyer, it has halved its staff numbers, though it is still making Nao (and a serving assistant on wheels called Plato). The uncertainty around the company's future has stoked fears that it will become impossible to get its robots repaired in future, and that Aldebaran could stop supporting the AI cloud network that the machines need to access to be able to function. What does this mean for the future of robotics research in universities? Besides fears about Aldebaran's future, there have long been issues with Pepper and Nao's durability. They both have rigid, fragile plastic shells, and the machines sometimes overheat. This means they have to be left to cool down after 20-30 minutes, which has often interfered with experiments and data-gathering -- as documented in this 2022 study of Nao. A spokesperson for Aldebaran agreed that motors can overheat, depending on their use and environment. They said the next generation of Nao, currently in development, has taken this into account in its design. For repairs, the only option is Aldebaran or an authorized reseller, or you risk voiding your warranty. This typically involves shipping overseas, which can be slow and costly -- more so if the replacement parts are out of stock. One of us (Emilia) encountered this during the COVID pandemic. Nao's batteries need to be used regularly to keep functioning, which led the university's machine to fail because it was inaccessible during lockdowns. Aldebaran couldn't supply replacement batteries quickly, which halted research projects at the university for many months and meant that important submission deadlines were missed. Meanwhile, software upgrades for Pepper stopped when the company halted development in 2021 (sales stopped in 2024). This robot's limited processing capabilities make it troublesome to run the large language models (LLMs) that power interfaces like ChatGPT (although these can be run in conjunction with a computer with modifications). Nao does have an AI edition that can handle LLMs, though this too requires external modifications. Nao's upgrades also seem to have been limited, which in our experience appears to have made them more error-prone too. Both robots are already considerably less useful for research purposes in our opinion. Finally, Nao and Pepper were not built with adaptability in mind. Unlike more recent machines like the 3D-printed InMoov, made by French designer Gael Langevin, there's no way of customizing their components or appearance. Their fixed expressions, gestures and plastic body make them difficult to adapt to different user needs or applications, such as helping at home or in healthcare. This again reduces their usefulness from a research point of view. Addressing these concerns, the Aldebaran spokesperson said, "Spare parts availability on Nao is very good, [barring] the normal supply chain issues, and these were exacerbated during COVID like the rest of the commercial world. Pepper is more limited as it has not been in production for some time, but we are generally able to solve any customer issues. "Nao is still very active as a product, with production continuing along with software upgrades. We recently launched Nao Activities, a major software upgrade that provides generative AI capabilities for Nao." The spokesperson added that there were no plans to switch off AI cloud support for Nao or Pepper, and that the robots are not difficult to use in robotics research, "testament of which is the thousands of units being used in that environment". What can be done? If Pepper and Nao do become unusable for research, universities will have to either scrap them or try to redevelop them with custom parts and components. It's possible they could be hacked and gutted, replacement parts could be 3D-printed, new microprocessors installed and the software made local and open source, which may be enough to get the robots back up and working again. However, it probably makes sense for researchers to look forwards instead. But towards what? At a time when university finances are very tight, there may be a reluctance to buy new machines with potentially limited shelf lives. Robots from alternative providers such as Futhat and Unitree are supported by similar cloud-based AI systems. Some institutions may consider reallocating vital funding to other departments, with a significant impact across robotics research and education. Universities are at the heart of robotics research, upholding high ethical standards and rigorously testing machines without the conflicts of interest that manufacturers can have. Universities can also bring together diverse disciplines like computer science, engineering and cognitive science, fostering collaboration that encourages innovation. With the UK number one globally for research quality in this field, these are the training grounds for the next generation of roboticists at a time when there is a growing skills shortage. A different way forward would be for universities to start building and programming robots from scratch. For the cost of a new research robot, say £15,000, you could buy several high-spec 3D printers, hardware and components. This wouldn't be about building entire humanoid robots but prototypes of key aspects such as facial expressiveness or skin, human gestures or emotions. This would allow students to gain important hands-on engineering and programming skills, while conducting novel research exploring current gaps in the field. It would make personalizing them easier and repairing them quicker and cheaper, if you could 3D-print parts or use parts that could be easily replaced off-the-shelf. If universities are to remain relevant in this rapidly evolving field, it's vital that they learn from their difficulties with Pepper and Nao. At a time when robots are starting to be perceived as reliable and cost-effective support for people, this is a cautionary tale for all.