The goal of physical AI is to enable engineers to program physical agents in a manner similar to how they handle digital ones. However, the field of robotics is currently hindered by a lack of data from real-world environments. Engineers often rely on mock-up spaces, while a growing sector focuses on monitoring factory floors and gig workers to gather data for training deep learning models for robotic functions.
Simulation presents a viable alternative, allowing the creation of detailed virtual replicas of real-world settings to generate necessary data for robotic development. Antioch, a startup specialising in simulation tools for robotics, aims to bridge the “sim-to-real gap,” ensuring that robots trained in virtual environments can perform reliably in the physical world. Co-founder and CEO Harry Mellsop underscores the need to make simulations indistinguishable from actual environments for autonomous systems.
Recently, Antioch secured an $8.5 million seed funding round, valuing the company at $60 million. Founded in May last year by Mellsop and four co-founders—including experienced professionals from companies like Transpose and Google DeepMind—the startup has the potential to transform how smaller companies develop robotics technologies without needing extensive capital for physical testing setups.
Antioch’s tools enable robot developers to create multiple digital instances of their hardware, simulating real-world sensor data for testing purposes. While the self-driving car industry, including companies like Waymo, has begun to utilise advanced simulations, there remains a significant opportunity to enhance the use of simulation in robotics.
The startup’s platform aims to support newer companies that lack the resources to build their own physical testing environments. As Mellsop states, the majority of the industry currently underutilises simulations, highlighting the urgent need for advancement in this area.
In comparisons with successful software tools, Antioch provides a platform where robots can be tested against various scenarios to refine their operations. However, achieving high-fidelity simulations that accurately reflect physical laws remains a challenge. The company is working with established models from companies like Nvidia to enhance its simulation library.
As the focus shifts towards sensor and perception systems, Antioch’s early partnerships with major multinationals showcase its position in the market. The need for robust simulation capabilities for developing safety standards and high-accuracy tasks is crucial, as highlighted by industry experts.
While physical AI advances, the development of tools for autonomous systems is still in its infancy. Mellsop predicts that in just a few years, autonomous systems will increasingly rely on software-driven solutions, enabling continuous iteration and improvement.
Currently, research projects at institutions like MIT are already exploring the potential of simulation tools in robot design. However, significant challenges still lie ahead in bridging the gap between digital models and real-world applications. For companies looking to emulate the successes of leaders in the field, investing in simulation tools or acquiring them will be essential.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
