Self-driving cars could soon possess human-like intuition after researchers developed an artificial intelligence system that predicts what pedestrians will do next rather than simply reacting to their movements.
Researchers at Texas A&M University and the Korea Advanced Institute of Science and Technology have unveiled “OmniPredict”, a breakthrough system that applies the technology behind advanced chatbots to read human behaviour on city streets.
While traditional autonomous systems struggle to adapt when people behave unexpectedly, this new model combines visual cues with contextual information to anticipate human actions in real time, effectively giving vehicles “street smarts”.
“Cities are unpredictable. Pedestrians can be unpredictable,” said Dr. Srikanth Saripalli, the project’s lead researcher and director of the Center for Autonomous Vehicles and Sensor Systems. “Our new model is a glimpse into a future where machines don’t just see what’s happening, they anticipate what humans are likely to do, too.”
Proactive prevention
The system marks a major shift from reactive autonomy to proactive prevention. Unlike standard computer-vision models that rely on brute-force visual learning, OmniPredict interprets the scene to understand motives — reading posture changes, hesitation and body orientation.
Early tests against the JAAD and WiDEVIEW datasets — two of the toughest benchmarks for pedestrian behaviour — showed the system achieving 67 per cent accuracy, outperforming the latest models by 10 per cent even without specialised training.
“Fewer tense standoffs. Fewer near-misses. Streets might even flow more freely. All because vehicles understand not only motion, but most importantly, motives,” said Saripalli.
The technology’s ability to read signs of stress or hesitation has implications far beyond traffic safety. Researchers believe the system could be deployed in military and emergency operations to detect threatening cues or provide an extra layer of situational awareness.
“For instance, the possibility of a machine to capably detect, recognise and predict outcomes of a person displaying threatening cues could have important implications,” said Saripalli. “Our goal in the project isn’t to replace humans, but to help augment them with a smarter partner.”
