Researchers from Princeton Engineering and the Indian Institute of Technology (IIT) have made significant progress in reducing the time and cost associated with wireless chip design by leveraging artificial intelligence (AI).
The AI design features unusual, and efficient, circuity patterns. Image Credit: Emir Ali Karahan, Princeton University.
Specialized microchips that enable cutting-edge wireless technology represent remarkable feats of engineering and miniaturization. However, designing these chips is both costly and time-intensive.
In a study published in Nature Communications, researchers detail a methodology where AI generates intricate electromagnetic structures and circuits within microchips based on specified design parameters. This process, which previously required weeks of expert effort, can now be completed in just hours.
Interestingly, the AI-driven approach has resulted in unconventional designs featuring unusual circuitry patterns. According to Kaushik Sengupta, the lead researcher, these designs often surpass the performance of traditional chips and include features that might not have been conceived by human designers.
We are coming up with structures that are complex and look random shaped and when connected with circuits, they create previously unachievable performance. Humans cannot really understand them, but they can work better.
Kaushik Sengupta, Professor and Study Lead Researcher, Department of Electrical and Computer Engineering, Princeton University
These AI-designed circuits can be optimized for energy efficiency or tailored to operate across an extensive frequency range, far beyond current capabilities. Furthermore, the new method synthesizes highly complex structures in minutes, compared to the weeks required by conventional algorithms. In some cases, the AI has even produced designs that current techniques cannot replicate.
Uday Khankhoje, a co-author and associate professor at IIT Madras, emphasized the groundbreaking nature of this work.
This work presents a compelling vision of the future. AI powers not just the acceleration of time-consuming electromagnetic simulations, but also enables exploration into a hitherto unexplored design space and delivers stunning high-performance devices that run counter to the usual rules of thumb and human intuition.
Uday Khankhoje, Associate Professor and Study Co-Author, Electrical Engineering, Indian Institute of Technology Madras
Wireless chips combine standard electronic circuits, like those found in computer chips, with electromagnetic components such as antennas, resonators, and signal splitters. These elements are painstakingly crafted and integrated into circuits, which are then scaled up to form complex systems. This traditional design process is labor-intensive, especially for advanced chips used in wireless communication, autonomous vehicles, radar systems, and gesture recognition.
Classical designs, carefully, put these circuits and electromagnetic elements together, piece by piece, so that the signal flows in the way we want it to flow in the chip. By changing those structures, we incorporate new properties. Before, we had a finite way of doing this, but now the options are much larger.
Kaushik Sengupta, Professor and Study Lead Researcher, Department of Electrical and Computer Engineering, Princeton University
The complexity of wireless chip design is almost incomprehensible. According to Sengupta, the number of potential configurations for a modern chip exceeds the number of atoms in the universe. Due to this overwhelming complexity, human designers typically take a bottom-up approach, adding components incrementally and fine-tuning the design as they progress.
AI, however, approaches the challenge differently. It views the chip as a single cohesive system, leading to innovative and unexpected arrangements. Despite its potential, AI still requires human oversight, as it can produce flawed designs alongside efficient ones.
“There are pitfalls that still require human designers to correct. The point is not to replace human designers with tools. The point is to enhance productivity with new tools. The human mind is best utilized to create or invent new things, and the more mundane, utilitarian work can be offloaded to these tools,” said Sengupta.
The research team has already used AI to design complex electromagnetic structures paired with circuits for broadband amplifiers. Moving forward, Sengupta plans to explore the design of entire wireless chips by connecting multiple structures using the AI system.
“Now that this has shown promise, there is a larger effort to think about more complicated systems and designs. This is just the tip of the iceberg in terms of what the future holds for the field,” said Sengupta.
Journal Reference:
Ali Karahan, E., et al. (2024) Deep-learning enabled the generalized inverse design of multi-port radio-frequency and sub-terahertz passives and integrated circuits. Nature Communications. doi.org/10.1038/s41467-024-54178-1