Parallax Worlds, a startup building hyper-realistic virtual simulations to stress-test robots before deployment, today announced it raised $4 million in a seed round.

Together with the company’s pre-seed funding, this brings the total raised to $4.9 million. Pear VC led the round, with participation from GS Futures, Kakao Ventures, and Lightscape Partners, alongside numerous angel investors and groups.

Developing and deploying robots can be a fraught process, especially in the current era of smart machines powered by artificial intelligence. Even simple robots require extensive on-site testing before they are ready for production lines.

“The biggest challenge that we’re trying to solve is to make sure that once we deploy these robots out in the real world, where failures matter a lot, you see fewer of those failure cases come up,” Tanmay Agarwal, co-founder and chief executive of Parallax Worlds, told SiliconANGLE in an interview. “Downtime at these factories is one of the biggest costs…especially as we’re beginning to think about AI-based robots going into these production systems.”

In fact, the cost of robot downtime can range substantially from $1,000 to a colossal $10,000 per minute, according to industry reports. This is tied primarily to the type of use case or product the robot is working with. Calculating the exact cost of downtime is difficult, but the downtime itself isn’t the only problem industrial and other applications face but costs generated by operational disruption.

Agarwal added that 80% to 90% of small and mid-sized manufacturing sites have not yet adopted a single robot due to deployment being “too expensive and too time-consuming.”

Parallax provides an integrated platform that allows developers to turn simple videos into accurate, interactive 3D environments. Using the platform, teams can quickly generate fully physics-enabled digital twins and run interactive simulations using real robot software.

The company’s founding team brings significant experience in both advanced robotics and industrial simulation. Agarwal trained in Stanford’s robotics community under renowned AI experts Fei-Fei Li and Jiajun Wu, developing software for platforms ranging from home robotics to self-driving vehicles. Aumkar Renavikar, co-founder and chief technology officer, studied robotics at the University of Illinois Urbana-Champaign and later worked on factory-floor automation at BMW AG and Michelin.

Parallax integrates with numerous industry physics engines and tooling — including Nvidia Corp.’s Omniverse, Unreal and Unity — while adding layers that simulate robot-specific behaviors and AI-assisted scenario generation.

Making quick virtual testing the norm

“You deploy one of these robots and when the bay door opens, the sun shines right through the door. Now the entire lighting of the scene has changed. Shadows, reflections [can cause] repeated failures that never came up in testing,” Agarwal said.

He added that simulation allows companies to test hundreds of edge cases — including variations in lighting, object placement and camera noise — that would otherwise be impractical physically.

For example, virtually recreated testing scenarios based on videos of actual production can add artificial lights, change the time of day, randomize the position of the robot, products and other factors. All of this allows companies to rapidly test events that robots might encounter.

Attempting to test even a few situations a robot might experience in real life, it can take years of trial and error and cost tens of thousands or more. Implementing robot control software in a simulation can reduce development time to weeks and significantly lower costs.

“Parallax is building the missing simulation layer that bridges robotics R&D and real-world deployment by virtualizing physical spaces to make testing and reliability scalable,” said Payam Banazadeh, visiting partner at Pear VC.

Real-world visibility in industrial use cases

The company boasts a number of robotics customers across manufacturing, construction and customer-service use cases, including industrial manufacturing robotics maker Orangewood Labs Inc., home construction robot firm BotBuilt Inc. and mobile welding robot developer Rainier Labs LLC.

“Parallax lets us validate robot behavior against real factory conditions using our actual control software,” said Abhinav Das, CEO of Orangewood Labs. “Being able to iterate on perception, planning and edge-case handling before hardware is deployed materially accelerates our development cycle.” 

Agarwal said the company’s future simulation capabilities will also be useful for solving potential safety issues by adding virtual people. That way, developers can test robots against the unpredictable behavior of human workers, including realistic walking patterns and other common interactions.

For example, with autonomous mobile robots commonly deployed in warehouses, the platform will be able to calculate near-miss distances in centimeters to validate compliance and safety standards.

The company’s short-term goals include automating the entire simulation pipeline using an AI agent that can generate test cases from natural-language instructions. This would allow Parallax to scale test-case generation dramatically, involving hundreds of cloud-based simulation runs for every deployment.

Meanwhile, Agarwal said the company intends to use the funding to expand beyond robotics startup customers to make inroads with manufacturers and warehouse operators. Using the company’s software, these industries could reduce robot-deployment timelines from a year to just a few months of assessment and testing.

To support this expansion, Parallax plans to increase its team by making additional hires and to enhance its research and development efforts.

Image: Pixabay