SpatioTemporal selected as finalist for Propel-AIR

Melbourne, VIC - Motion intelligence company SpatioTemporal has been selected as one of seven finalists in Propel-AIR 2026, Australia's dedicated AI and robotics sprint delivered by the ARM Hub AI Adopt Centre.

The company will now compete for the opportunity to work with and alongside MassRobotics in Boston, the global epicentre of robotics innovation, with its motion intelligence software for robots and autonomous vehicles operating around people.
 

Teaching machines to read the room
Today's autonomy systems detect objects well, but struggle with human intent. Who goes first at a doorway? Is that lean the start of a step, or a weight shift? As robots and autonomous vehicles move out of the lab and into shared space, these judgement calls will happen more and more frequently. Getting them wrong means hesitation at best, and collision at worst.
 

"Humans are not structured environments," says SpatioTemporal founder and CEO Andrew Ballard. "We are variable, social and occasionally inconvenient. But that variability is not noise. It carries information." SpatioTemporal builds a software layer that sits between a robot's perception and planning systems, predicting what the people and objects around it are about to do.
 

"Perception can tell a robot there are five people in a room. Planning can chart a route between them," Ballard said. "But only Motion Intelligence can help it understand who is stepping aside, who is approaching, who has changed their mind and who the robot should pay attention to."

Rather than analysing pixels or waypoints, the company's foundation model treats movement as a language, modelling object kinematics as 10,000 discrete motion tokens and reading intent from the transitions between them.
 

Half a million encounters in simulation
SpatioTemporal has run more than half a million simulated robot-human proximity encounters in NVIDIA Cosmos, a widely used AI stack for autonomy. In the company's most difficult scenario, the baseline system recorded near-collisions in 24 per cent of runs. Adding SpatioTemporal's motion intelligence signals cut that to 2 per cent, with robots yielding earlier, moving more smoothly and making fewer unstable changes of direction.
 

The potential market spans every machine that works near people: humanoid robots, autonomous vehicles and warehouse systems. SpatioTemporal's advisors include Professor Alan Duffy, Pro Vice-Chancellor for Flagship Initiatives at Swinburne University, connecting the company into Australia's research ecosystem as it scales.
 

Growth plans
SpatioTemporal is raising a pre-seed round to fund a deliberately lean nine-month sprint: prove its motion intelligence on one live use case, formalise intent-aware benchmarks, and convert that proof into partnership and licensing conversations across the AI, robotics and automotive sectors.
 

The winner will travel to MassRobotics in Boston for a one-month residency, with access to leading robotics companies and world-class institutions like MIT and Harvard.
 

ENDS

About SpatioTemporal 
SpatioTemporal builds motion intelligence for physical AI: a software layer between perception and planning that predicts the intent of people and objects in shared spaces. Its foundation model, LSTM-01, treats movement as a language, enabling robots and autonomous vehicles to anticipate what will happen next. The company was founded by Andrew Ballard.
 

About Propel-AIR 
ARM Hub's Propel-AIR program is Australia's AI and Robotics Sprint, designed to fast-track Australian robotics and AI innovators by equipping them with tools, guidance, and networks to scale their business. This pioneering program provides business 
opportunities for AI and robotics companies, culminating in a virtual pitch to an international panel of experts. The winner receives a month's residency in Boston with MassRobotics.

Media Contact: Mike Woodcock, Communications Director, ARM Hub
M: 0411 969 248 
E: Mike.Woodcock@armhub.com.au 
W: armhub.com.au