MIT Unveils 6-Milliwatt Gleanmer Chip for Real-Time 3D Mapping, Cutting Power Use to 2.5%
Updated
Updated · MIT News · Jun 23
MIT Unveils 6-Milliwatt Gleanmer Chip for Real-Time 3D Mapping, Cutting Power Use to 2.5%
1 articles · Updated · MIT News · Jun 23
Summary
Gleanmer generated real-time 3D maps using about 6 milliwatts of power in MIT tests, making detailed onboard mapping feasible for tiny drones, robots and other battery-limited devices.
The chip cuts energy use by pairing dedicated hardware with MIT’s GMMap algorithm, which builds maps from compact Gaussian shapes instead of voxel grids and processes depth images in a single pass.
That design lets the system keep only a few pixels and recently seen objects in fast on-chip memory, avoiding repeated image storage and reducing the need for power-hungry off-chip data transfers.
MIT said Gleanmer uses about 2.5% of the power required by the best existing mapping chip, and path planning on the resulting maps needs roughly 20% of the energy otherwise required.
Researchers see uses beyond industrial inspection drones, including lightweight AR headsets, and plan to push efficiency further by moving processing closer to the sensors.
Will this tiny, power-sipping chip finally make all-day augmented reality glasses practical?
Could this 3D mapping chip become the eyes for the next generation of embodied AI robots?
Gleanmer: The First 6 mW SoC for Real-Time 3D Gaussian Occupancy Mapping at the Edge
Overview
On June 23, 2026, MIT researchers unveiled Gleanmer, a groundbreaking System-on-a-Chip (SoC) designed for real-time 3D Gaussian Occupancy Mapping. Gleanmer stands out for its extremely low power consumption of just 6 milliwatts, making it ideal for detailed environmental mapping on battery-limited devices. This efficiency is achieved through a sophisticated design that includes a specialized accelerator for GMMap and advanced algorithm-hardware co-optimizations, such as direct computation and efficient reuse of compact Gaussians. As a result, Gleanmer can cut map construction energy by up to 63%, marking a new era for low-power 3D mapping.