Answers

8.2 Collecting data

Learning · Robot club · about 20 min

What this lesson is about

Readings with labels, and why bad data makes bad robots.

• Four situations
• Getting to the spots
• Bad data makes bad robots

Questions 6 marks in all

1. [1 mark]In this module, what is a sample?

   1. AA view (the 16 level readings) paired with a label
   2. BJust the 16 level readings
   3. CThe label the robot guessed
   4. DThe robot's position when it recorded
   Answer: A. A sample is the numbers plus a word for what they mean. The model learns the link between the two.

2. [1 mark]Which of these produce a robot that does the wrong thing with total confidence?

   Tick every answer that is true.

   1. ARecording the wall once with the label open
   2. BRecording open only at the start
   3. CRecording three headings at each spot
   4. DUsing labels that mean where the robot is, not what it sees
   5. ERecording after the robot has settled
   Answer: A, B, D. Wrong labels, one place only, and labels that mean where instead of what. Recording from more headings and after settling are the fixes.

3. [1 mark]Why does drive_to end with wait(0.4) before a sample is recorded?

   1. ASo the readings are taken after the robot has settled, not while it is still sliding
   2. BThe depth sensor needs 0.4 s to switch on
   3. CTo give you time to type the label
   4. Dtof_grid() only works after a wait
   Answer: A. A view taken while sliding is a mixture of places. Settle first, then record.

4. [1 mark]What does this program print?

   data = []
   data.append(([54, 52, 51], "open"))
   data.append(([16, 16, 16], "wall-ahead"))
   print(len(data), data[-1][1], data[0][0][1])

   Answer:

   2 wall-ahead 52

   Two samples. data[-1][1] is the last sample's label; data[0][0][1] is the second reading of the first sample.

5. [1 mark]Who decides the label on a sample?

   1. AYou do, and if you get it wrong the model learns your mistake
   2. BThe robot works it out from the readings
   3. CThe depth sensor measures it
   4. DThe model fixes wrong labels during training
   Answer: A. The label is your judgement. The model can only be as right as the labels you give it.

6. [1 mark]What does the label gap-left mean?

   1. AThe wall ends on my left, wherever I happen to be
   2. BThe robot is standing on the left spot, (-38, 45)
   3. CTurn left now
   4. DThere is a gap somewhere on the left of the mat
   Answer: A. Labels describe what the robot sees, not where it is. The same view anywhere should get the same label.

The task: collect data

# the two lines every program starts with: the commands, then the robot
from bugbot import *
connect()
# maths: atan2, hypot, sin, cos, radians
import math

def wrapped(h):
    return (h + 180) % 360 - 180

def go_to(x, y, speed=60):
    # where am I?
    px, py = position()
    a = math.radians(wrapped(math.degrees(math.atan2(x - px, y - py)) - heading()))
    # forward, sideways, rotation: -100 to 100 each, until the next command
    drive(speed * math.cos(a), speed * math.sin(a), wrapped(0 - heading()) * 3)

def near(x, y, cm=4):
    # where am I?
    px, py = position()
    return math.hypot(x - px, y - py) < cm

def drive_to(x, y):
    while not near(x, y):
        go_to(x, y)
        # pause 0.1 s (the robot keeps doing what it was told)
        wait(0.1)
    # all motors off
    stop()
    # pause 0.4 s (the robot keeps doing what it was told)
    wait(0.4)
drive_to(0, 10)
print('recorded', 'open')

from bugbot import *
connect()
import math

def wrapped(h):
    return (h + 180) % 360 - 180

def go_to(x, y, speed=60):
    px, py = position()
    a = math.radians(wrapped(math.degrees(math.atan2(x - px, y - py)) - heading()))
    drive(speed * math.cos(a), speed * math.sin(a), wrapped(0 - heading()) * 3)

def near(x, y, cm=4):
    px, py = position()
    return math.hypot(x - px, y - py) < cm

def drive_to(x, y):
    while not near(x, y):
        go_to(x, y)
        wait(0.1)
    stop()
    wait(0.4)
data = []
spots = [("open", 0, 10), ("wall-ahead", 0, 45), ("gap-left", -38, 45), ("gap-right", 38, 45)]
for label, x, y in spots:
    drive_to(x, y)
    data.append((tof_grid()[16:32], label))
    print("recorded", label)
print("samples:", len(data))