New Mcity Digital Twin by Parallel Domain Brings Physical AI Research Closer to Reality

Parallel Domain today announced the completion of a digital Replica of the Mcity Test Facility at the University of Michigan. This high-fidelity simulation environment allows Mcity members and researchers to run perception, planning, prediction, and control tests with realism nearly indistinguishable from the physical track.

The Mcity PD Replica provides the visual fidelity required for accurate perception testing. Other digital twins are better suited for validating behavior and control systems, rather than the sensors and perception algorithms that are critical to ensuring safe autonomy. The Mcity PD Replica provides pixel-level visual detail, matched lighting, and accurate physics that make it possible to test the entire autonomous vehicle stack in an open or closed-loop simulation.

Built entirely from drive-log data captured by the Mcity team, this is one of the largest and most detailed PD Replica locations ever created. The PD Replica simulation, combined with the real Mcity track, offers teams a unique opportunity to cross-validate results, comparing simulation runs directly with on-track testing to ensure simulated performance mirrors reality.

From foundational digital twins to real-fidelity

Mcity’s first open-source digital twin made the track accessible to researchers worldwide. Subsequent releases brought physics-based sensor simulation to the platform. Now PD Replica Sim brings photorealism to camera, lidar, and radar sensor simulation unlocking perception-level validation in addition to control and behavior testing.

“The Mcity PD Replica represents a major leap forward in how our members and research partners can use simulation,” said Derek Johnson, Principal Engineer at Mcity. “By capturing the track in such extraordinary visual and physical detail, we’re enabling perception and control testing that was previously only possible on-site. And the beauty of a digital twin is that validation can still be done on-site in a real environment that matches the simulated environment.”

“Mcity has become more than a physical track, and we’re proud to help extend it into the digital domain,” said Kevin McNamara, Founder and CEO of Parallel Domain. “Our ability to turn Mcity’s drive logs into its most realistic simulation yet shows the future of Physical AI: developed and refined in simulation, accessible anytime, anywhere.”

Accelerating safe autonomy with realism that scales

In the race to deliver safe and reliable autonomous systems, real-world testing alone is not enough. It can take millions of on-road miles to encounter just one critical safety event, making physical development slow, costly, and high-risk.

Parallel Domain accelerates physical AI testing with photorealistic digital twins purpose-built for scalable validation. The platform enables both open and closed-loop testing in highly realistic, dynamic virtual environments that mirror the complexity of the real world.

By bridging the gap between simulation and reality, Parallel Domain creates a Software Augmented Testing environment, combining real-world testing and digital twins built from real-world data, drive logs, and sensor captures. This approach moves critical testing, quality checks, and performance monitoring earlier in the development cycle, allowing issues to be caught and resolved sooner.

Benefits of Software Augmented Testing

• Protect autonomous programs from catastrophic failure by ensuring integrity at every stage of the development cycle.
• Accelerate time-to-market by running thousands of simulations in parallel.
• Lower costs by shifting left to run scalable simulations earlier in development and reducing reliance on costly, slow real-world testing.
• Enhance safety and reliability by routinely testing AI systems against critical, rare, and exhaustive scenarios across thousands of variations in real PD Replica environments, providing added assurance that systems perform as intended
  

Drive the Mcity track anytime, anywhere

The Mcity PD Replica is available now to Mcity members, academic partners, and Parallel Domain customers via the PD Replica Sim API. For researchers and engineers, this means the Mcity track can be accessed remotely from anywhere in the world with the realism and fidelity needed for perception-level testing, no gas required.

About Parallel Domain

Parallel Domain provides API-driven simulation that turns real locations into real-fidelity virtual grounds for physical AI testing. Its customers in automotive, robotics, and aerospace use PD Replica Sim to accelerate perception development, regression testing, and safety validation.

About Mcity

Mcity is an interdisciplinary public-private partnership bringing together industry, government, and academia as part of the University of Michigan Transportation Research Institute. With a world-class test facility, mixed-reality testing capabilities, a data engine, and renowned talent, Mcity is a makerspace for research, development and deployment of next-generation mobility innovations to the real world.