Summary Points
- MIT’s new low-power chip enables tiny robots to create real-time 3D maps.
- It uses Gaussian ellipsoids for efficient obstacle and free space mapping.
- The chip consumes only 6 milliwatts, ideal for extended wearable AR devices.
- Innovative algorithms reduce memory use and power by processing compact Gaussians.
MIT has developed a new chip that helps tiny robots avoid obstacles inside complex environments. This chip is designed to allow small, low-power drones and robots to create detailed 3D maps in real time. They can do this using just as much power as a single LED, making them ideal for battery-limited devices. The chip can help robots navigate tight corners in places like industrial HVAC systems to detect gas leaks safely.
Innovative Mapping with Gaussians
Instead of using traditional cube-shaped pixels called voxels, the MIT team used ellipsoid-shaped blobs called Gaussians. These blobs more accurately match curved surfaces and save space by representing large areas with fewer shapes. The system uses an efficient algorithm called GMMap to generate 3D maps quickly from depth images. It compares only neighboring pixels, reducing memory and power needs. The map is made even more compact by merging overlapping Gaussians, which is done without going back to the original images.
Powerful and Compact Hardware Design
The system-on-a-chip, called Gleanmer, accelerates the process with specialized hardware. It keeps the active Gaussians in small on-chip memory, making the process faster and more energy-efficient. Tests show it can reconstruct 3D environments while consuming only 6 milliwatts of power—about 2.5% of what other chips need. This low power use enables small robots and devices like augmented reality headsets to operate longer without recharging. Open future applications include better blueprints analysis and enhanced environmental sensing, helping robots and AR devices navigate environments seamlessly.
Stay Ahead with the Latest Tech Trends
Explore the future of technology with our detailed insights on Artificial Intelligence.
Access comprehensive resources on technology by visiting Wikipedia.
QuantumV1
