Controlling a Sailboat with

an Artificial Neural Network

Ibrahim
21060116130099
Sistem Kontrol Cerdas

Source: Alex Likos Paper, Engineering and Computer Science Union College
Point of Sailing

A = No-go Zone
B = Close-hauled
C = Beam Reach
D = Broad Reach
E = Running
Part of a Sailboat

 Keel
Wind Indicator  Sail
 Main Sheet
 Rudder
 Tiller
 Wind Indicator

Sail

Tiller

Keel Main Sheet Rudder

Back Propagation Procces

1 2 3 4 5 6

1. Present a training sample to the neural network

2. Compare the network’s output to the desired output for that
input. Calculate the error in each output neuron.
3. For each neuron, calculate what the output should have been,
and a scaling factor, how much lower or higher the output must
be adjusted to match the desired output. This is the local error.
4. Adjust the weights of each neuron to lower the local error.
5. Assign “blame” for the local error to neurons at the previous
level, giving greater responsibility to neurons connected by
stronger weights.
6. Repeat from step 3.
 Input Output

Input :
• Boat facing
• Wind Speed Output:
• Wind Direction • Position of the
• How much the OUTPUT sheet (the ropes
boat is heeling that one uses to
• How the skipper is INPUT control the sail)
distributing his • The position of
weight to rudder
counteract the
heeling of boat
• The location of a
waypoint
Neural Network

Sail Simulator 5
(wind speed,
a direction etc with
NMEA format)

b σ

c Ɵ

f
 Neural sailboat steering

Sea State
Neural
Neural Procces
Controller
Network P
(1/P) Output
Learning - Learning -
Input

+ +

Desired
1/P
Respons
 Neural Network Test
2000 iterations:
0.004
 0.004 error of rudder position
0.002
 0.000002 error of sail position
8000 iterations:
 0.002 error of rudder position
0.000002
0.0002  0.0000005 error of sail position
10000 iterations:
0.0000005
 0.0002 error of rudder position
0.0000002  0.0000002 error of sail position

2000 8000 10000

Although the project does not deal specifically with the real world, it is using the
same kind of information that is obtainable from digital sensors.
A neural network that was trained using data generated by realistic simulator will
perform just fine in the real world.