Singapore's $100 million center to train humanoid robots for bomb scenes, prison searches, and hazmat missions
Humanoid robots for public safety
How do you train humanoid robots for public safety missions? If you're Singapore, you build a massive $100 million center to train them for four different missions that Singapore thinks humanoids will play a huge role in not too many years in the future.
Here are Singapore's physical AI training goals:
- Mission one: hazardous-material containment, mitigation, and sampling
- Mission two: post-blast scene investigation and analysis
- Mission three: prison cell searches for contraband
- Mission four: restocking medical supplies in ambulances
These are among the most dangerous tasks Singapore's Home Team officers perform.
So on September 30, the Home Team Science and Technology Agency opened a $100 million, 29,000-square-foot facility dedicated to training humanoid robots to handle all of them. The Home Team Humanoid Robotics Centre — H2RC — is described as the world's first facility exclusively focused on public safety humanoid robotics.
The deployment target: a humanoid robot working alongside officers in a real operational setting by 2028.
We first noted Singapore's efforts in this area in a Humanoid (almost) Daily newsletter about a week ago. Now, more details are available.
Why these 4 missions
Most conversations about humanoid robots focus on warehouses, factories, and eventually, homes. The applications are high-volume and the economics are relatively legible: replace or augment labor in controlled environments, measure throughput, calculate ROI.
Singapore is working on a different problem.
Its Home Team — the collective name for the Singapore Civil Defence Force, the Singapore Police Force, and the Singapore Prison Service — regularly sends people into situations where the danger is high, the environment is unstructured, and the consequences of failure are severe. HAZMAT containment requires entering contaminated zones. Post-blast investigation means entering buildings that may still be structurally compromised, with debris and uncertain hazards. Prison searches expose officers to resistance and hidden contraband. Ambulance restocking is time-critical work at the edge of a medical response.
Each mission puts a human at risk. A humanoid robot that can do these tasks doesn't need to match human performance on day one — it just needs to be safer to use than a person in those specific scenarios.
That's a different bar than "beat the warehouse packer on ROI."
What the center actually is
H2RC's 29,000 square feet include mock public safety environments: realistic post-blast debris scenes, industrial infrastructure replicas with pipes and valves, and purpose-built zones for data collection and AI model training. The facility houses a high-performance computing infrastructure to run the AI development work.
The most distinctive piece of kit is an omnidirectional multi-robot overhead mobility system — the first of its kind in Southeast Asia. Previously, that technology was used by the European Space Agency for astronaut training. H2RC has repurposed it to give robots controlled movement and orientation during training scenarios.
The training philosophy is explicit: the center is designed to transition robots from preprogrammed, scripted behaviors toward embodied AI that can learn, reason, and respond in dynamic environments. Preprogrammed systems can handle known sequences in known environments. The scenarios H2RC is building for don't offer those guarantees.
At launch, the center demonstrated 4 operational humanoid robots — 1 for each mission type. The police robot can kneel for ground-level post-blast work. The prison search robot is equipped with dexterous hands and tactile sensing for probing surfaces and detecting hidden objects. These aren't generic platforms borrowed from a manufacturer's warehouse; they're purpose-built for the specific physical demands of each mission.
HTX also signed an MOU with ST Engineering for expanded cooperation in humanoid robotics. ST Engineering is a Singapore defense contractor with a global presence in aerospace, smart city, and defense technology — a partnership that signals both operational intent and the kind of institutional backing that turns pilot programs into production deployments.
The timeline and what it means
"Our goal is clear: we aim to have the first humanoid robot working alongside Home Team officers in a real operational setting, performing defined tasks safely and reliably by 2028," said Tan Chye Hee, Permanent Secretary of Home Affairs.
That's a pointed thing to say. "Performing defined tasks safely and reliably" is a much more demanding standard than "completing demos successfully." It implies operational reliability, failure mode management, and the kind of safety validation that most humanoid companies are still working toward.
Singapore is giving itself 2 years.
The 2027 interim target — assisting with firefighting and rescue missions — provides a midpoint check. If robots can contribute to SCDF operations in 2027, the 2028 goal of full alongside-officer deployment isn't aspirational. It's a logical next step.
What this means beyond Singapore
The humanoid industry has been arguing about where the first major commercial deployments will happen. Manufacturing has the most momentum. Logistics is the most legible. Home use is the biggest long-term prize.
Public safety sits outside that conversation almost entirely, and Singapore just put $100 million on it.
The implications are meaningful. Public safety robots don't need to hit cost-per-labor-hour benchmarks. They need to do things humans can't safely do, or do them without putting people at risk. That's a purchase criteria that bypasses most of the ROI arguments that currently make humanoid economics complicated.
If H2RC's robots work — if a humanoid can effectively search a prison cell, investigate a blast scene, or enter a hazmat environment — the proof of concept travels. Every government with dangerous public safety missions has an interest in what Singapore learns.
That's a large potential customer base. And it's one that has historically been willing to pay for capability without waiting for the consumer market to bring costs down.
Singapore is a small country that has consistently made large bets on technology before it was obvious. H2RC may be the most consequential thing anyone has done in humanoid robotics this month.
Sources
Written by
HMH