Home · Module 1 · Robotics & AI
The Sense → Think → Act loop
Every robot — from a Roomba to Atlas — runs the same fundamental cycle. AI changes how rich each stage can be, not the loop itself.
Click each stage below to see what happens inside, and what AI techniques can replace or enhance the classical approach.
Click any stage
Each of the three nodes opens a deep-dive into what that stage actually does, the classical engineering approach, and how AI/ML can supercharge it.
Loop frequency matters. A factory arm runs its loop at 1000 Hz. A Roomba at 50 Hz. A self-driving car at 10–100 Hz depending on the stage. Slower loops → more time to think → can use bigger models. Faster loops → tighter control → must be simpler.
Where AI sits in the loop
Traditionally, robotics engineers hand-coded each stage. Modern robotics uses ML/AI for the parts that are hard to specify rules for:
| Stage | Classical | AI / ML |
|---|---|---|
| Sense | Edge detection, Hough transforms, thresholding | CNNs for object detection, segmentation, depth estimation |
| Think | A* search, behavior trees, state machines | Reinforcement learning policies, LLM planners, MPC |
| Act | PID controllers, inverse kinematics solvers | Learned dynamics models, neural feedback control |
Your Raspberry Pi car as the loop
- Sense: HC-SR04 ultrasonic returns distance in cm.
- Think:
if distance < 20: turn()— start with this. Upgrade to a learned policy later. - Act: PWM signal to L298N motor driver → wheels turn.
- Loop: Repeat ~10 times per second in your Python script.