Why the robot thinks for itself.
Four reasons
- Speed. On-robot answers take milliseconds. Cloud calls take seconds. In MoLogica testing the trained inspection model answered in about 1 ms on the robot's GPU against about 3 s for a cloud call.
- Resilience. Mines and remote sites lose connectivity. A robot that needs the cloud to see is a robot that stops.
- Privacy. Some sites won't let footage leave. Local models keep it there.
- Cost. Every on-robot answer is one you don't pay a cloud API for.
How it gets there
Large cloud models judge first and produce labels. Small models learn from those labels and move onto the robot. The cloud stays as teacher and judge.
Read more: data and learning.
CPU and GPU in one paragraph
A CPU has a few powerful cores that do one complex thing after another: logic, coordination, step-by-step decisions. A GPU has thousands of small cores doing the same simple maths on huge amounts of data at once: images, AI models, rendering. On a robot the GPU looks and recognises and the CPU decides and coordinates. The Jetson puts both on one board with one pool of memory.
On-robot language models
Open models have been shown running on an entry-level Jetson fast enough to take spoken or written instructions on the robot with no cloud. On-robot language models are on the MoLogica roadmap.
On-robot inspection inference is proven in simulation; on-robot language models are planned.
Ready to see robots patrolling your site?
Before any hardware is bought, we can build a simulated copy of your site and show the fleet patrolling it at night while you configure the checks in plain language.