Agility Robotics has unveiled the next version of its humanoid robot Digit. Digit 5 is the company's first model engineered to work safely near humans.
The point here is less what the robot can do than where it is allowed to stand. Until now robots in this class worked inside separate cells behind physical barriers.
How it behaves
When Digit 5 detects a person at a distance it can take precautions on its own. It can move to avoid the person, or stand still so they can pass by without getting closer.
When someone does come into close proximity, a different response kicks in: the robot can squat and assume a seated position. That single move lowers both the risk of toppling and the height of any impact.
| Detected situation | The robot's response |
|---|---|
| Person at a distance | Changes course |
| Person wanting to pass | Stands still and waits |
| Person coming close | Squats and sits |
A system, not a single rule
According to Pras Velagapudi, chief technology officer at Agility, the robot "was designed with a complex safe motion system that can take a variety of different mitigations depending on exactly what sort of human presence is detected".
That distinction matters. A single "stop when you see a human" rule makes a robot unusable in a busy warehouse. Giving different responses to different distances preserves safety while letting the work flow.
This graded approach also departs from the established logic of industrial robotics. Classic safety design rested on separating the robot from the human; here the aim is not separation but managing the possibility of contact.
Why the fence question matters
Safety fences are not merely a cost item but a constraint on layout. Setting up a robot cell demands floor space and a redesign of the workflow around it.
Remove the fence and the robot can step into the line where people are working. That is the company's claim: safety features of this kind could unlock far more opportunities in warehouses and automotive factories, without isolated robot work cells and physical separation barriers.
The part still to be proven
The claim is strong, but verification happens in the field. Safety systems in this class generally have to pass independent certification, and those processes take longer than laboratory demonstrations.
The second question is robustness: in how many of ten runs does the detection work correctly. A warehouse has dust, changing light and crowds all at once, and the value of a safety feature rests on its success rate under exactly those conditions.
A third question is cost. The saving from removing a fence has to be weighed against the expense of fitting the robot with more complex sensing hardware, and neither figure is public yet.
What it changes on the floor
Logistics and automotive are the areas where humanoid robots are expected to arrive first. Fence-free operation means existing warehouses could take on robots without being redesigned, which directly changes the investment calculation.
On the other hand, workplace safety regulation looks not at a robot's capability but at its certification. What decides matters for a facility is therefore not that the robot can squat, but that the behaviour has been approved against the relevant standard.