Researchers at Shanghai’s Tongji University have developed a first-generation specialized computing chip designed to enhance rolling optimization capabilities for autonomous vehicles, robots, drones, and other intelligent machines. The device, named the Moving Horizon Unit, functions as an advanced processing core that allows machines to observe their environment, analyze information, and adjust behavior in real time.

Rolling optimization—a process by which systems continuously update decisions and actions based on changing environmental feedback—is a natural human ability, explained Chen Hong, a lead scientist and professor at Tongji University’s College of Electronic and Information Engineering. Conventional general-purpose chips, Chen noted, often struggle to manage the dynamic decision-making required by autonomous systems operating in complex and variable settings.

The research team’s objective was to translate this adaptive capability into dedicated hardware that empowers machines to respond proactively without relying strictly on predefined programming. Their chip integrates a specialized computing architecture that encompasses innovations such as predictive modeling of future conditions directly within the hardware, allowing for efficient option evaluation. Additionally, it merges principles of artificial intelligence with physical laws, enabling devices to learn and adapt while minimizing the need for extensive training data.

Based on three decades of experience in combining algorithms with hardware, the team holds exclusive intellectual property rights over this architecture. Early trials have included the development of an onboard controller and real-vehicle integration tests, demonstrating the chip’s functionality and real-world application potential.

The Moving Horizon Unit’s core module can be tailored for diverse needs and incorporated into various chip systems, thus broadening its applicability across multiple sectors. At the chip’s unveiling on May 29, Frank Allgower, former president of the International Federation of Automatic Control and a professor at the University of Stuttgart, emphasized that the technology addresses longstanding challenges related to the high computational demands of rolling optimization. He highlighted that the chip reduces these burdens and simplifies implementation, potentially enabling uses in scenarios previously considered difficult.

Demonstrations at Tongji University’s intelligent and connected vehicle testing base featured chip-equipped vehicles performing complex maneuvers such as serpentine driving and lane changes across different environments, underscoring the chip’s operational viability.

Looking ahead, the research team intends to focus on iterative improvements, deepen collaborations with industry partners, and accelerate deployment across fields including intelligent manufacturing, industrial control, new energy, and robotics.