Cambridge Startup Vsim Launched Robot Training System

The robotics firm developed a virtual simulation tool to help machines anticipate and adapt to unexpected scenarios.

Updated on Sept. 21, 2026 in Robotics

Bold flat-color editorial illustration of a geometric robotic arm with orange accents, representing virtual simulation technology. — Cambridge-based startup Vsim launched a new robotics training system that uses local simulation to improve machine adaptability in unpredictable real-world scenarios. AI Illustration. Upload story photo >

Live Poll

Do you trust that emerging robotics technology is currently safe for use in homes and offices?

Cambridge-based startup Vsim introduced a training system that utilizes virtual environments to prepare robots for physical tasks. By performing millions of simulations, the software enables robots to identify optimal solutions and improve adaptability.

Why it matters

This technology allows robots to anticipate potential outcomes and react to unpredictable events in real time. It marks a significant step in minimizing the gap between simulated training and real-world execution.

The system is optimized for graphics processing units and runs on hardware carried directly by the robots. It processes millions of task iterations in computer simulations to refine motion and decision-making.

The players

Vsim

This is a Cambridge-based robotics startup founded by Michelle Lu and Kier Storey.

Nvidia

This major technology firm provides a robotics software suite and a world model known as Cosmos.

Rika Antonova

She serves as an associate professor at the Cambridge Department of Computer Science and Technology.

Michelle Lu

She is one of the co-founders of the Cambridge startup Vsim.

Kier Storey

He is one of the co-founders of the Cambridge startup Vsim.

The details

The software, which is optimized for graphics processing units, runs locally on the robots to simulate movement scenarios. By iterating through millions of simulations, the system ensures robots can handle unexpected situations without manual intervention.

September 21, 2026: The startup introduced its robotics training software.

The Tech Race

This development reflects the broader industry shift toward utilizing advanced world models like the Nvidia Cosmos to bridge the reality gap in robotics. It positions small, specialized firms as key partners in the ongoing arms race to automate complex physical tasks.

As these systems improve, robots will likely become more capable of operating safely in complex, unpredictable environments like homes and workplaces. Users can expect more reliable automation in machines that rely on these advanced training simulations.

The takeaway

The move toward high-fidelity simulation suggests that the future of robotics lies in software that can predict outcomes before a machine ever moves. Developers should prioritize training algorithms that emphasize adaptability to ensure systems remain functional outside of controlled lab environments.