pizz.dev

Research

A paper on AI for tiny robots that run on a $10 chip, and the start of a new line: neuromorphic computing, AI that works more like the brain. The supercomputing work is with the projects.

Papers

The key figure of the Nano-U paper
  1. Nano-U: Efficient Terrain Segmentation for Tiny Robot Navigation

    Federico Pizzolato, Francesco Pasti, Nicola Bellotto

    Towards Autonomous Robotic Systems (TAROS) 2026, accepted

Research lines

Left: a garden path as seen by the robot camera. Right: the same picture with the part Nano-U marks as safe to drive on highlighted in yellow.

Bachelor's thesis, paper at TAROS 2026 · 2025In progress

Nano-U

A tiny AI model that lets a small robot recognise the ground it can drive on, using a camera and a $10 chip with 320 KB of memory. It is 700 times smaller than the model it learned from, with almost the same accuracy.

A first experiment in neuromorphic computing

Research, just started · 2026In progress

Neuromorphic computing

The start of my next research line: AI that works more like the brain, with spiking neurons that only compute when something changes, to run on very little energy.