Researchers from the United States and the Netherlands have shown that controlled randomness in the movement path of robots reduces traffic jams and improves the rate of task completion, without the need for a powerful central computer or complex coordination.
When you put more and more robots in the same workspace, it seems at first glance that productivity should only increase. More robots mean more “working hands,” and therefore faster completion of tasks such as cleaning a contaminated area, transporting equipment, or assembling a machine. But new research shows that at some point the exact opposite happens: the crowding turns traffic into a traffic jam, the robots interfere with each other, and overall productivity begins to decline. According to the study, the solution is not necessarily more sophisticated central control, but rather adding a precise amount of randomness to each robot’s path (seas.harvard.edu).
The study, published on February 17, 2026 inProceedings of the National Academy of Sciences (PNAS) , was carried out by Lucy Liu, Justin Werfel, Federico Toshi and L. Mahadvan. The researchers combined mathematical analysis, computer simulations and physical experiments with small robots to examine how dense groups of agents move towards fixed goals in a shared space. According to the scientific abstract, they showed that above a critical level of disturbance, large and persistent traffic jams cease to exist, and that it is possible to mathematically calculate the level of density and the level of randomness that result in the maximum number of goals reached per unit time. (PubMed)
Fewer straight lines, more flow
In the model the researchers built, each robot was given a random starting location and a random destination. Once the robot reached its destination, it was immediately assigned a new destination, creating an environment that simulated a fleet of robots performing a sequence of tasks. When the robots moved in perfectly straight lines, that is, without any interference, they tended to get stuck in each other and create dense congestion zones. When too much randomness was introduced, they stopped getting stuck in traffic jams, but they began to zigzag excessively and waste time. The best result appeared precisely in the middle area, where there was enough deviation from the straight line to elude each other, but not to the point of losing direction.
Lucy Liu, who led the study, explained that the result sounds counterintuitive: On the face of it, randomness should make prediction more difficult, not better. But when there is an appropriate amount of randomness, it is possible to calculate stable averages of distances, times, and movement patterns, and thus better predict collective behavior. The researchers used this insight to develop formulas that estimate the “goal achievement rate” of the entire group, not just an individual robot.
Not just a simulation, but also a laboratory experiment
To make sure the idea didn’t just stay at the mathematical model level, the team collaborated with Eindhoven University of Technology in the Netherlands. There, swarms of small, wheeled robots were set up, and the researchers tracked their movements using an overhead camera. Each robot was attached with a QR code, allowing the system to track its location and assign it new goals. Although the real robots were slower and clumsier than the agents in the simulation, the key patterns repeated themselves: a little randomness improved flow, too much randomness hurt efficiency, and a lack of randomness created blockages.
One of the important conclusions of the study is that it is not necessary to run a central supercomputer or to give each robot a particularly high level of “intelligence” to achieve effective coordination. According to the researchers, at least at certain densities, a set of simple local rules can suffice. The abstract of the article itself also states that a simple reactive navigation method worked well up to moderate density levels, and was even much more computationally efficient than a sophisticated central planner.
This conclusion may have broader implications than industrial robotics. The researchers note that the principles revealed here may also help in the design of dense public spaces, in understanding pedestrian flows, in the coordination of autonomous vehicles, and even in the study of “active matter” and collective behavior in the living world. In other words, the question of how a dense group of units moves without getting stuck in each other is relevant not only to robot swarms, but also to cities, roads, ports, warehouses and emergency centers.
The central message of the study is simple but important: An efficient system is not always one that tries to impose perfect order. Sometimes it is precisely a small, controlled, and calculated deviation from the straight path that prevents paralysis, releases the bottleneck, and allows the entire group to work better. In a world where autonomous robots are expected to operate increasingly in shared spaces, this is an insight that may transform from a theoretical generality into a fundamental design principle.
More of the topic in Hayadan:
- Sheepdogs reveal a better way to guide robot swarms
- At virtual reality school, fish teach robots
- A Breakthrough in Collaborative Robotics: The "Curvature Criterion" Inspired by Ant Swarms
- Flight of the Robotic Bees / Robert Wood, Radhika Nagpal and Gu-Yeon Wei
- NASA is considering sending swimming robots to "ocean worlds" that might be suitable for life
