SINGAPORE: When Mr Tommy Ng posts a job opening for coffee shop cleaners, it can take weeks to get responses. When he does hire someone, they often do not stay long.
"The turnover is very high because of the tedious work. It could be every few months there will be a change," said the business development manager for Mr Teh Tarik Eating House. The five coffee shops the company runs each require at least two cleaners per shift.
The struggle to hire and retain cleaners is a common refrain among coffee shop operators, who said the physically demanding and "unglamorous" nature of the work makes it increasingly difficult to attract and keep people on the job.
"The barrier of entry is not high, so (workers have) the mindset that if they do not work here today, they can work elsewhere and they will not cherish the job as much," said Mr Wu Yi Feng, director of Kopiwu, which runs coffee shops in Ang Mo Kio, Clementi and Yishun.
While the mandatory tray return rule introduced several years ago has eased workloads to some extent, staff are still needed to man return points, wipe tables and keep dining areas clean.
Mr Andy Hoon, CEO of the Kim San Leng coffee shop chain, pointed to Singapore's foreign manpower quotas as a factor further restricting the pool of available workers. Operators also said they prefer to hire cleaners directly rather than through cleaning companies to avoid higher operating costs.
With human workers in short supply, some operators have turned to automation. The experiments span nearly a decade – from tray-return robots introduced in 2016 to table-cleaning robotic arms still in development.
The latest attempt is perhaps the most unusual: a table that cleans itself.
At Nanyang Technological University's (NTU) Pioneer Food Court, operator Food Haven has installed a prototype that replaces its tablecloth at the push of two buttons.
The four-seater table uses disposable sheets stretched across its surface. When activated, a mechanism pulls the soiled sheet – along with food scraps and spills – into a compartment at one end while dispensing a fresh layer from a roll at the other end. The process takes about 30 seconds.
Each roll contains 30 sheets, enough for three to four days of use before cleaners must replace it. A tablet connected to the table tracks how many sheets remain.
Food Haven director Shane Tay said his team developed the concept in response to staffing difficulties. The food court, which seats about 300 diners, currently requires three to four cleaners during peak hours. He hopes the technology could eventually cut that number in half and ease the cleaners' workload.
"There's always this struggle of maintaining proper hygiene within the large dining area of food courts such as this because of manpower issues," Mr Tay said. "It's hard to hire workers and hard to retain workers also."
"We are trying to eliminate this dependency on manpower (with the automation of the cleaning process). So every diner after ... (they) are done eating, they return their trays. The next group of diner can just sit down with the press of buttons."
Each table currently costs between S$500 and S$800 to produce, though Mr Tay said mass manufacturing could lower the price. If the trial succeeds, Food Haven may consider upgrades such as battery operation and more sustainable materials.
This NTU food court table can swap out a synthetic sheet at the push of two buttons, offering each diner a clean surface. (Photo: CNA/Tan Wen Lin)
Food scraps and spilt liquid are pulled into a built-in compartment of the self-cleaning table. (Photo: CNA/Tan Wen Lin)
At the opposite end of the table, a built-in compartment contains a roll of sheets that will provide diners with a clean surface. (Photo: CNA/Tan Wen Lin)
Food scraps and spilt liquid on a disposable sheet of a self-cleaning table being piloted at NTU on Nov 29, 2025. (Photo: CNA/Tan Wen Lin)
Director of Food Haven Shane Tay at the self-cleaning table he helped design. (Photo: CNA/Tan Wen Lin)
When CNA tested the table, larger food scraps often became stuck at the narrow gap leading to the collection compartment. Some liquids also spilt inside, requiring manual cleaning afterwards.
The table's bulky ends reduce seating capacity, and several diners initially mistook it for a hotpot or barbecue setup. Most patrons were unaware of its self-cleaning function.
NTU students said dirty tables are common during peak lunch hours, with food scraps, drink condensation and occasional bird droppings seen at the open-air premises. Some said they bring wet wipes to clean tables themselves before eating.
Undergraduate Chua Jia Xing said the design could improve hygiene without requiring cleaners to intervene, though she raised concerns about waste generation and tripping hazards from the power cable.
Spilt liquid sometimes ends up in the built-in compartment. (Photo: CNA/Tan Wen Lin)
Food Haven cleaner Mdm Vasantha said the table was simple to operate and required less cleaning than conventional tables, which helps lighten her workload.
Other operators expressed interest but raised questions about reliability and cost.
"This sort of technology, at times it will break down. So how long will the downtime be? Will it affect operational needs?" asked Mr Teh Tarik Eating House’s Mr Ng. "At the end of the day if ... such technology spoils, then we still need a person to go there and clean up the table."
Kim San Leng's Mr Hoon found the concept promising but wondered whether used rolls might produce odours after several hours. Kopiwu's Mr Wu questioned the environmental impact and noted that his coffee shops use both rectangular and round tables, making a fixed design harder to implement.
The open-air food court in Nanyang Technological University (NTU) on Nov 29, 2025. (Photo: CNA/Tan Wen Lin)
Mdm Vasantha, a cleaner at Food Haven food court in Nanyang Technological University, said the self-cleaning table has helped to ease her workload. Apart from cleaning tables, Mdm Vasantha sorts returned crockery at the food court's tray return stations, as seen on Dec 5, 2025. (Photo: CNA/Tan Wen Lin)
Food Haven's table is not the first attempt to automate cleaning at Singapore's food centres. Other technological solutions have been tested over the years, with limited success.
Tray-return robots were introduced at coffee shops and hawker centres as early as 2016, to help diners return used crockery before tray return became mandatory. But diners and cleaners complained that the robots often obstructed walkways and still required cleaning. When mandatory tray return was introduced, fixed return points made the robots redundant.
Food centres also previously tested table-cleaning robots, including a robotic arm equipped with cloths and a bin that could move to dirty tables and wipe surfaces. According to a 2022 press release from Weston Robot, the company behind the technology, the machine could detect items on tables before cleaning.
Asked for an update, founder Zhang Yanliang said the firm had identified areas for improvement and indicated the robot would undergo further research and development.
A robot arm developed by Weston Robot for table cleaning being tried out at what appears to be a hawker centre. (Screenshot: YouTube/Weston Robot)
Experts said current robotics technology cannot yet replicate how humans clean at food centres because of the dynamic nature of these environments, where flexibility and judgment are essential.
Associate Professor Harold Soh from the National University of Singapore's computer science department described food centre environments as a “super challenge” for robots, which must contend with different patrons, objects and unexpected incidents.
“Effective cleaning in such a space requires both physical intelligence (dexterous hands, balance) and social intelligence (knowing when to move, when to stop, and how not to annoy or injure customers)," he said. “Robots today struggle with both.”
While robots can run, jump or do backflips in controlled settings, they frequently fail at everyday tasks like picking up irregular objects, handling soft or slippery food waste, or stacking items reliably, he added.
“Dexterous manipulation remains one of the most challenging problems in robotics.”
Robots must also plan their actions while predicting how they affect their surroundings – a combination of perception, prediction and communication that remains "an unsolved problem outside of controlled environments", Assoc Prof Soh added.
Associate Professor Lyu Chen from NTU's School of Mechanical and Aerospace Engineering agreed that robots still struggle with "complex, fast-changing public settings" where fine manipulation and judgment are needed.
Most commercial robotic systems still rely heavily on pre-programmed routines, he said. While navigation and perception have improved, table cleaning remains difficult because seemingly simple actions involve complex reasoning, such as determining whether a table is vacant, identifying leftovers versus valuables and recognising different types of spills.
“In public spaces, human-robot interaction is not optional. The system has to behave safely, smoothly and efficiently around people, otherwise it will not be usable at scale,” he said.
He noted that automation can handle routine tasks like crockery collection or back-end transport, freeing cleaners to focus on areas requiring judgment like spills and toilets. But human supervision remains necessary even for partially automated systems.
Tray-return robots that were once mobile are now stationary tray-return points at a coffee shop and food court. (Photos: CNA/Koh Wan Ting)
Operators said they have yet to find suitable technology that can be deployed at their premises, and even with automation, manpower remains essential.
Yet they see no easy way around the labour crunch. Mr Wu pointed to Singapore's ageing population, while Mr Ng noted that most cleaners are retirees, a pool that is both limited and increasingly unable to handle physically demanding work.
Dr Ng Boon Yuen, a senior lecturer at Singapore University of Social Sciences' School of Business, said technological solutions are more likely to succeed when intended users are involved in the design, selection and implementation.
If adoption is low at food centres, it's worth reconsidering if the technology has been designed based on the actual needs of the users, she added.
Assoc Prof Soh said researchers are working on integrating AI with robots that can interact with the real world – a field known as embodied AI. Key areas include equipping robots with multiple sensors for richer environmental understanding, and improving their reasoning and planning capabilities.
"The science is advancing, but getting robots out of the lab and into everyday coffee shops remains a challenge," he said.
Continue reading...
"The turnover is very high because of the tedious work. It could be every few months there will be a change," said the business development manager for Mr Teh Tarik Eating House. The five coffee shops the company runs each require at least two cleaners per shift.
The struggle to hire and retain cleaners is a common refrain among coffee shop operators, who said the physically demanding and "unglamorous" nature of the work makes it increasingly difficult to attract and keep people on the job.
"The barrier of entry is not high, so (workers have) the mindset that if they do not work here today, they can work elsewhere and they will not cherish the job as much," said Mr Wu Yi Feng, director of Kopiwu, which runs coffee shops in Ang Mo Kio, Clementi and Yishun.
While the mandatory tray return rule introduced several years ago has eased workloads to some extent, staff are still needed to man return points, wipe tables and keep dining areas clean.
Mr Andy Hoon, CEO of the Kim San Leng coffee shop chain, pointed to Singapore's foreign manpower quotas as a factor further restricting the pool of available workers. Operators also said they prefer to hire cleaners directly rather than through cleaning companies to avoid higher operating costs.
CAN TECHNOLOGY FILL THE GAP?
With human workers in short supply, some operators have turned to automation. The experiments span nearly a decade – from tray-return robots introduced in 2016 to table-cleaning robotic arms still in development.
The latest attempt is perhaps the most unusual: a table that cleans itself.
At Nanyang Technological University's (NTU) Pioneer Food Court, operator Food Haven has installed a prototype that replaces its tablecloth at the push of two buttons.
The four-seater table uses disposable sheets stretched across its surface. When activated, a mechanism pulls the soiled sheet – along with food scraps and spills – into a compartment at one end while dispensing a fresh layer from a roll at the other end. The process takes about 30 seconds.
Each roll contains 30 sheets, enough for three to four days of use before cleaners must replace it. A tablet connected to the table tracks how many sheets remain.
Food Haven director Shane Tay said his team developed the concept in response to staffing difficulties. The food court, which seats about 300 diners, currently requires three to four cleaners during peak hours. He hopes the technology could eventually cut that number in half and ease the cleaners' workload.
"There's always this struggle of maintaining proper hygiene within the large dining area of food courts such as this because of manpower issues," Mr Tay said. "It's hard to hire workers and hard to retain workers also."
"We are trying to eliminate this dependency on manpower (with the automation of the cleaning process). So every diner after ... (they) are done eating, they return their trays. The next group of diner can just sit down with the press of buttons."
Each table currently costs between S$500 and S$800 to produce, though Mr Tay said mass manufacturing could lower the price. If the trial succeeds, Food Haven may consider upgrades such as battery operation and more sustainable materials.
This NTU food court table can swap out a synthetic sheet at the push of two buttons, offering each diner a clean surface. (Photo: CNA/Tan Wen Lin)
Food scraps and spilt liquid are pulled into a built-in compartment of the self-cleaning table. (Photo: CNA/Tan Wen Lin)
At the opposite end of the table, a built-in compartment contains a roll of sheets that will provide diners with a clean surface. (Photo: CNA/Tan Wen Lin)
Food scraps and spilt liquid on a disposable sheet of a self-cleaning table being piloted at NTU on Nov 29, 2025. (Photo: CNA/Tan Wen Lin)
Director of Food Haven Shane Tay at the self-cleaning table he helped design. (Photo: CNA/Tan Wen Lin)
PRACTICAL LIMITATIONS
When CNA tested the table, larger food scraps often became stuck at the narrow gap leading to the collection compartment. Some liquids also spilt inside, requiring manual cleaning afterwards.
The table's bulky ends reduce seating capacity, and several diners initially mistook it for a hotpot or barbecue setup. Most patrons were unaware of its self-cleaning function.
NTU students said dirty tables are common during peak lunch hours, with food scraps, drink condensation and occasional bird droppings seen at the open-air premises. Some said they bring wet wipes to clean tables themselves before eating.
Undergraduate Chua Jia Xing said the design could improve hygiene without requiring cleaners to intervene, though she raised concerns about waste generation and tripping hazards from the power cable.
Spilt liquid sometimes ends up in the built-in compartment. (Photo: CNA/Tan Wen Lin)
Food Haven cleaner Mdm Vasantha said the table was simple to operate and required less cleaning than conventional tables, which helps lighten her workload.
Other operators expressed interest but raised questions about reliability and cost.
"This sort of technology, at times it will break down. So how long will the downtime be? Will it affect operational needs?" asked Mr Teh Tarik Eating House’s Mr Ng. "At the end of the day if ... such technology spoils, then we still need a person to go there and clean up the table."
Kim San Leng's Mr Hoon found the concept promising but wondered whether used rolls might produce odours after several hours. Kopiwu's Mr Wu questioned the environmental impact and noted that his coffee shops use both rectangular and round tables, making a fixed design harder to implement.
The open-air food court in Nanyang Technological University (NTU) on Nov 29, 2025. (Photo: CNA/Tan Wen Lin)
Mdm Vasantha, a cleaner at Food Haven food court in Nanyang Technological University, said the self-cleaning table has helped to ease her workload. Apart from cleaning tables, Mdm Vasantha sorts returned crockery at the food court's tray return stations, as seen on Dec 5, 2025. (Photo: CNA/Tan Wen Lin)
OTHER ATTEMPTS AT AUTOMATION
Food Haven's table is not the first attempt to automate cleaning at Singapore's food centres. Other technological solutions have been tested over the years, with limited success.
Tray-return robots were introduced at coffee shops and hawker centres as early as 2016, to help diners return used crockery before tray return became mandatory. But diners and cleaners complained that the robots often obstructed walkways and still required cleaning. When mandatory tray return was introduced, fixed return points made the robots redundant.
Food centres also previously tested table-cleaning robots, including a robotic arm equipped with cloths and a bin that could move to dirty tables and wipe surfaces. According to a 2022 press release from Weston Robot, the company behind the technology, the machine could detect items on tables before cleaning.
Asked for an update, founder Zhang Yanliang said the firm had identified areas for improvement and indicated the robot would undergo further research and development.
A robot arm developed by Weston Robot for table cleaning being tried out at what appears to be a hawker centre. (Screenshot: YouTube/Weston Robot)
Related:
TECHNOLOGY NOT THERE YET: EXPERTS
Experts said current robotics technology cannot yet replicate how humans clean at food centres because of the dynamic nature of these environments, where flexibility and judgment are essential.
Associate Professor Harold Soh from the National University of Singapore's computer science department described food centre environments as a “super challenge” for robots, which must contend with different patrons, objects and unexpected incidents.
“Effective cleaning in such a space requires both physical intelligence (dexterous hands, balance) and social intelligence (knowing when to move, when to stop, and how not to annoy or injure customers)," he said. “Robots today struggle with both.”
While robots can run, jump or do backflips in controlled settings, they frequently fail at everyday tasks like picking up irregular objects, handling soft or slippery food waste, or stacking items reliably, he added.
“Dexterous manipulation remains one of the most challenging problems in robotics.”
Robots must also plan their actions while predicting how they affect their surroundings – a combination of perception, prediction and communication that remains "an unsolved problem outside of controlled environments", Assoc Prof Soh added.
Associate Professor Lyu Chen from NTU's School of Mechanical and Aerospace Engineering agreed that robots still struggle with "complex, fast-changing public settings" where fine manipulation and judgment are needed.
Most commercial robotic systems still rely heavily on pre-programmed routines, he said. While navigation and perception have improved, table cleaning remains difficult because seemingly simple actions involve complex reasoning, such as determining whether a table is vacant, identifying leftovers versus valuables and recognising different types of spills.
“In public spaces, human-robot interaction is not optional. The system has to behave safely, smoothly and efficiently around people, otherwise it will not be usable at scale,” he said.
He noted that automation can handle routine tasks like crockery collection or back-end transport, freeing cleaners to focus on areas requiring judgment like spills and toilets. But human supervision remains necessary even for partially automated systems.
Tray-return robots that were once mobile are now stationary tray-return points at a coffee shop and food court. (Photos: CNA/Koh Wan Ting)
NO EASY SOLUTION
Operators said they have yet to find suitable technology that can be deployed at their premises, and even with automation, manpower remains essential.
Yet they see no easy way around the labour crunch. Mr Wu pointed to Singapore's ageing population, while Mr Ng noted that most cleaners are retirees, a pool that is both limited and increasingly unable to handle physically demanding work.
Dr Ng Boon Yuen, a senior lecturer at Singapore University of Social Sciences' School of Business, said technological solutions are more likely to succeed when intended users are involved in the design, selection and implementation.
If adoption is low at food centres, it's worth reconsidering if the technology has been designed based on the actual needs of the users, she added.
Assoc Prof Soh said researchers are working on integrating AI with robots that can interact with the real world – a field known as embodied AI. Key areas include equipping robots with multiple sensors for richer environmental understanding, and improving their reasoning and planning capabilities.
"The science is advancing, but getting robots out of the lab and into everyday coffee shops remains a challenge," he said.
Continue reading...