Traffic jams to power AI as scientists transform roads into computers

Researchers at Tohoku University in Japan have developed a novel AI approach called Harvested Reservoir Computing (HRC) that leverages real-world traffic dynamics as a computational resource, significantly reducing energy consumption compared to traditional AI systems. This method builds on reservoir computing principles by using the complex, naturally occurring interactions within urban road networks—specifically traffic flow—as a "reservoir" to process data. The team, led by Professor Hiroyasu Ando, tested this concept through a combination of scaled autonomous car experiments and simulations of grid-like urban traffic, discovering that predictive accuracy peaks at a critical medium-density traffic state just before congestion occurs. This state offers the most diverse and informative traffic dynamics, enabling accurate forecasting with minimal computational overhead. Importantly, the HRC approach does not require new specialized hardware; it can utilize existing traffic sensors and observational data, making it practical for real-world application. The researchers propose that roads and other social infrastructures can be viewed as continuously operating computers, potentially transforming smart city management