Scalable Surface-Based Manipulation Through Modularity
The problem
Robotic Manipulation Surfaces (RMS) move objects by deforming the surface they rest on. This makes them attractive for fragile, irregular, or food-like objects, where grasping can damage the item or require a specialized gripper. The difficulty is scale: dense actuator arrays can give fine control, but they become expensive and mechanically complex when the workspace grows.
The core idea
The paper asks whether a manipulation surface can be made modular. Instead of building one very dense surface, we connect fabric-based modules and coordinate object movement across module boundaries. Adjacent modules share edge actuators, reducing actuator count from 4n^2 to (n + 1)^2 for an n x n grid.
In a 2 x 2 prototype, this gives a 1 x 1 m workspace with only 9 actuators.
What makes it hard
Sharing actuators is efficient, but it mechanically couples neighboring modules. Moving one module can disturb objects on another module. The platform therefore needs both high-level path planning and low-level compensation.
Our framework combines:
- Manhattan-style conflict-free path planning.
- Directional object passing between modules.
- A geometric PID controller for positioning.
- Compensation strategies for passive-object disturbance.
Result
The system achieves coordinated inter-module transfer and sub-centimeter positioning for heterogeneous objects, including fragile items. Compensation reduced passive-object displacement by 59-78% across the measured configurations.
Why I care about this
For me, the interesting part is not only that a surface can move objects. It is that manipulation can be treated as a property of a distributed surface: hardware design, geometry, and control all contribute. That feels like a promising route toward scalable manipulation in logistics, food processing, and other settings where objects are too varied for one gripper.
Paper: arXiv:2601.21884