SpatioTemporal
[Adjective] Relating to both space and time.
SpatioTemporal is an Australian research company building foundation models for Physical AI.
Founded in Melbourne, we develop compact foundation models and operating systems for robotics – with research spanning motion, intent, human-robot interactions and consequence modelling.
Our first foundation model, LSTM-01, established Motion Intelligence as the first Large SpatioTemporal Model. We are now expanding the technical programme through simulation, benchmarking, multi-agent research and collaboration with robotics and Physical AI partners.
Physical AI is moving from controlled environments into a world shaped by people.
Machines can increasingly perceive that world and plan actions within it. The harder problem is interpreting how it changes.
Humans learn this remarkably early. We read hesitation, commitment, convergence and risk from movement, often before an action becomes explicit.
SpatioTemporal was founded around a simple proposition: movement is information, and machines should be able to read it too.
We believe that capability belongs between perception and planning as a distinct layer of Physical AI.
Research with a path to deployment.
SpatioTemporal sits at the intersection of machine learning, robotics and human behaviour. We are deliberately not building robots, vehicles or autonomy stacks. We build intelligence that can integrate with them.
Our current work combines foundational research with practical evaluation: developing models, testing them in simulation, establishing benchmarks, and working with partners to understand where Motion Intelligence can improve real autonomous systems.
TEAM
AB – Andrew Ballard
Matt Dickinson
Professor Alan Duffy
Andrew is the founder of SpatioTemporal and architect of its Motion Intelligence research. His background spans robotics, engineering, spatial systems, human-machine teaming and applied Defence AI research. He developed the LSTM-01 architecture and leads SpatioTemporal’s research, technical strategy and demonstrations.
Matt is SpatioTemporal’s COO and operating partner, with senior delivery and commercial experience across major Australian telecommunications and utility organisations. He leads commercialisation, partnerships, operations and investor process, with a focus on moving SpatioTemporal’s research into practical evaluation and integration.
Alan is an astrophysicist, science communicator and Pro Vice-Chancellor at Swinburne University of Technology. He advises SpatioTemporal on scientific strategy, research framing and institutional collaboration, bringing deep experience across research, technology and public science.
How we work
Robotics
Evaluate Motion Intelligence alongside existing perception and planning systems.
Simulation
Test shared-space interactions and generate richer variation in human behaviour.
Research
Develop benchmarks, datasets and evaluation methods for motion and intent understanding.
Physical AI infrastructure
Explore integration, inference and deployment pathways for compact foundation models.
Connect