MIT researchers enable self-assembling robots to locate themselves without GPS
Robots now build structures while tracking their own positions using only touch sensors—no cameras or GPS needed.
Researchers from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) have developed a new localization method for modular robots that enables self-assembling structures to determine their positions without relying on external tracking systems or expensive sensors. Published on arXiv (arXiv:2608.02895), the study introduces a contact-driven localization framework that uses only binary contact information—whether two robots are physically connected—to iteratively refine the pose of each module.
The approach leverages a virtual-force model where robots attract toward dock-connected neighbors and repel from non-connected ones, effectively simulating physical forces to achieve accurate positioning. The method was validated in simulations for assembling complex structures like towers and cantilevers, demonstrating scalability and precision without external infrastructure such as cameras or GPS.
- Modular robots use only onboard touch sensors for localization, eliminating the need for external tracking infrastructure.
- The virtual-force framework enables iterative pose refinement using binary contact data (connected/not connected).
- Simulations successfully demonstrated accurate self-assembly of freeform structures like towers and cantilevers.
Why It Matters
Enables scalable, low-cost robotic swarms for construction, search-and-rescue, and space exploration without GPS or cameras.