Rolling optimization chip enhances smart machine capabilities

Researchers from Shanghai-based Tongji University have unveiled their independently developed first-generation specialized computing chip for rolling optimization, called the Moving Horizon Unit, which functions like an intelligent brain for autonomous vehicles, robots, drones and other smart machines.

The research team said the chip enables intelligent devices to observe their surroundings, process information and adjust their actions in real time. The technology is expected to support the development of smart transportation and advanced intelligent equipment.

Chen Hong, a lead scientist on the project, said rolling optimization is a capability humans use naturally, continuously adjusting actions based on changing conditions and feedback to achieve a goal.

In complex and constantly changing environments, autonomous vehicles, robots and drones must continuously detect changes around them, plan their next moves and make decisions in real time. These tasks can be difficult for conventional general-purpose chips.

"Our research aims to transform the human capability of adapting to environmental changes and uncertainties in a proactive way into a machine-computable hardware solution," said Chen, a professor at Tongji University's College of Electronic and Information Engineering.

"By embedding this rolling optimization mechanism into a specialized chip, we enable vehicles to drive autonomously and robots to make independent decisions without relying on pre-set programs," he said.

Drawing on three decades of experience in integrating algorithms with hardware, the team developed a specialized computing architecture for rolling optimization and holds fully independent intellectual property rights to the technology.

According to Chen, the architecture includes several innovations. One allows the chip to predict possible future conditions and evaluate different options directly within the hardware, helping devices make decisions more efficiently. Another combines artificial intelligence with the laws of physics, enabling machines to learn and adapt with less dependence on massive amounts of training data.

"We have already developed an onboard controller and conducted real-vehicle integration tests," Chen said.

Researchers said the chip's core module can be customized for specific applications while also being integrated into a variety of chip systems, enabling it to be adapted for different uses.

At the chip's launch event on May 29, Frank Allgower, former president of the International Federation of Automatic Control, said rolling optimization has long been recognized for its advantages but has been difficult to apply widely because it requires substantial computing power.

"This new chip reduces these burdens, simplifying implementation and opening up possibilities for previously challenging systems and scenarios," said Allgower, who is also a professor at the University of Stuttgart in Germany.

Researchers said the chip has successfully passed real-world tests in intelligent vehicle control, evidencing its technical feasibility.

The team plans to continue upgrading the chip, expand joint research and development with industry partners and accelerate its use in fields including intelligent manufacturing, industrial control, new energy and robotics.

Attendees at the launch event also watched demonstrations at the university's testing and evaluation base for intelligent and connected vehicles. During the demonstrations, vehicles equipped with the chip performed maneuvers including serpentine driving and lane changes in a range of scenarios.