PYTHON · NUMPY

Perception becomes a decision

Close the loop: turn the classifier's output into an action so the agent reaches its goal.

01Challenge

Perception is only useful if it drives behaviour. The Bench gives a 1D agent, a perceive() that classifies the situation (goal behind, ahead, or reached), and a think() that ignores it and never moves. Map each perceived class to a velocity and close the loop. This is sense -> think -> act with a real perceptual front end -- the shape of every reactive robot.

02Model

A policy is a function from what you perceive to what you do. Here the perception is a class label and the policy is a lookup: behind means reverse, ahead means advance, reached means stop. Trivial as it is, this is the same wiring that turns a vision module's detections into a robot's motion.

03Guided practice

Step A (worked): think returns 0 for every class, so the agent sits at the origin -- the FAIL. Step B (the fix): return a velocity per class, e.g. {0:-1, 1:+1, 2:0}[cls]. Step C: the agent advances while the goal is ahead and stops when it arrives. Step D (connect): this is the block graph of the live sense-think-act lab, with a classifier as the sense block.

04Feedback

PASS WHEN PASS when the class-to-action map drives the agent to within 0.3 of the goal.

If your run fails, check:

Read the bench’s printed result — it passes when the class-to-action map drives the agent to within 0.3 of the goal. Run the cells top to bottom; if one errors, the FAIL line names the step to fix.
The agent never moves -- map class 1 to a positive velocity, class 0 to a negative one, class 2 to zero.

05Retrieve & space

What is a policy, in one sentence?
Why does the loop need the 'reached' class at all?
What plays the role of perceive() on a real robot?

06Mastery & project

Pass the lab: perception drives action and the agent arrives. You have wired the think block -- the third rung.

This is the sense-think-act loop made literal; the live lab lets you swap the hand-written think block for a learned one.

GO FURTHER · THE LIVE PLAYGROUND

See it as a graph: Sense to Think to Act ↗ — the same loop, where the think block can become a learned policy.

← A kilobyte classifier Clone the expert →

PYTHON · NUMPY

Perception becomes a decision

Close the loop: turn the classifier's output into an action so the agent reaches its goal.

You read

the arrays and values already in scope

You change

the code you write in each cell

Fixed

the dataset and the checks that grade you