The case study approach

In the invented Corvale scenario, a robot moving at 1.5 metres per second needs to stop for a child who steps out. A round trip to the cloud over the mobile network takes about 200 ms and sometimes fails. Why is pedestrian detection run on the robot's own computer rather than in the cloud?

A Corvale robot's ultrasonic sensors report distances in centimetres. Readings below 0 or above 400 are faults and must be ignored. What does this program print?

readings = [120, -1, 95, 900, 88]
closest = 400
valid = 0
for r in readings:
    if r >= 0 and r <= 400:
        valid = valid + 1
        if r < closest:
            closest = r
print(valid, closest)

Each Corvale robot has a stop button that a pedestrian may press at any moment, though it is pressed only a few times a month. The robot runs on a battery. How should the robot's software detect a press?

Corvale's object detector was trained mostly on daytime images from a much larger city. After dark, the robots in Corvale miss more pedestrians than in daylight. Which explanation fits best?

Corvale's cloud platform sends each one-time lid code to the robot over the mobile network. Which measure stops someone who intercepts the traffic from reading the code?

Tarnwick Health Centre's model was trained on 50,000 past messages, each already placed in the urgent, routine or admin queue by staff. Which type of machine learning is this?