Traditionally, robots prevent objects from slipping by simply increasing grip strength—a strategy that can damage fragile items and lacks finesse. A new study from the University of Surrey and collaborating institutions (including the University of Lincoln, Arizona State University, KAIST, and Toshiba Europe) offers a more human-like solution: predictive trajectory modulation, tells Tech Xplore.
Drawing inspiration from how humans subtly adjust hand movements—like tilting or slowing down when something begins to slip—the researchers engineered a robotic controller with a learned tactile forward model. This model anticipates slippage based on planned movements and triggers real-time adjustments in trajectory—not just grip force—to secure objects.
This bio-inspired method empowers robots to perform more intelligent manipulation, even when increasing grip strength is not feasible, such as with delicate, slippery, or asymmetric objects. Moreover, the system generalizes well: it works across unfamiliar objects and movement paths not seen during training, showcasing its versatility in real-world applications like manufacturing, health care, logistics, and domestic assistance.
In essence, this approach marks a shift toward more nuanced, dexterous, and adaptive robot manipulation—moving beyond brute force toward human-inspired control strategies that enhance both safety and reliability.