A team of researchers from the Technical University of Munich (TUM) has developed an innovative wheeled robot named Jack that seamlessly weaves through crowds. Unlike conventional robots that either stop abruptly or avoid interaction entirely, Jack predicts human movements and adapts accordingly to find the fastest and safest route.
With advanced technology, including a lidar system that creates a 360-degree map and sensors that measure speed and distance, Jack processes its surroundings 10 times per second. This allows it to anticipate how people around it will move and react.
TUM Professor Angela Schoellig explains, “Our robot models human reactions to plan its paths, setting it apart from typical systems that overlook interaction.”
Jack’s intelligence stems from a mathematical model inspired by human behavior. It learns to navigate like a human would, avoiding unnecessary stops while adjusting for changes in people’s paths.
For instance, if someone approaches Jack, it assumes they’ll move aside slightly—just as they might with another person. But if they stay on a collision course, Jack quickly recalculates and picks a different route.
The researchers use crowd behavior data to help the robot, Jack, learn and improve. Tested outside the lab, Jack continuously observes people and adjusts its routes, making its navigation more human-like.
Currently, Jack represents the third evolution in robot navigation. Beyond reacting to situations (stage 1) and predicting movements (stage 2), Jack is interactive (stage 3). It forecasts movements and influences human reactions while steering clear of collisions.
“Jack observes, predicts, and optimizes in real-time—almost like a human being,” says researcher Sepehr Samavi.
This breakthrough has implications far beyond crowd navigation. The researchers believe similar algorithms could enhance autonomous driving by enabling vehicles to anticipate and respond to human behavior on the road. Though humanoid robots may need extra work, delivery bots and mobility aids like wheelchairs could immediately benefit from Jack’s technology.
Professor Schoellig notes, “A moving robot can stop if needed, but humanoids are still too unstable to handle sudden changes.”
Despite this limitation, Jack’s success signals a promising future for robots that interact harmoniously with humans, bringing us closer to smarter, safer autonomous systems.
Journal Reference
- Sepehr Samavi, James R. Han, Florian Shkurti, and Angela P. Schoellig; SICNav: Safe and Interactive Crowd Navigation Using Model Predictive Control and Bilevel Optimization; IEEE Transaction on Robotics; 2025; DOI: 10.1109/TRO.2024.3484634
Source: Tech Explorist
