Edge AI: running AI on the robot itself
INTERMEDIATE · 6 MIN READ
Small robots cannot always call the cloud. Edge AI runs the model on the robot's own chip.
Cloud AI sends data to a big computer far away and waits for the answer. That adds delay and needs internet. Edge AI runs the model on a small chip on the robot.
To fit, models are made smaller and faster, using tricks such as quantisation (using fewer bits per number). You trade a little accuracy for a lot of speed and battery life.
A drone spots a tree 3 m ahead. It needs about 1.2 m to swerve. The AI answer takes time to arrive. How far does the drone fly before it can react?
Delay: 30 ms → flies 0.24 m before reacting · Time to swerve.
Try the cloud at 8 m/s with a 250 ms network. Then switch to on-board. Same AI, very different result. The trade-off: a small chip can run only a small model.
A drone avoiding trees cannot wait half a second for a cloud reply. It runs its vision model on-board.
Fast reactions, no internet needed, better privacy, and lower running cost.
Delays or lost connection make the robot freeze or act on old information.
Count how long a simple decision takes you to make. Now add one second of delay. Imagine driving like that.
Common questions
Can a Raspberry Pi run AI?
Yes, for small models such as simple object detection, often with a helper chip for more speed.
Is edge always better?
No. Heavy tasks can still use the cloud when delay does not matter.
Keep learning
Get your personal invite link to share R2BOT with friends.
Teachers and bloggers: paste the embed code into your page to show this demo for free.
Plain English. Free. Unsubscribe in one click.