carCalibration

Table of Contents

1. Offline Table

• Throttle Model
• Brake Model

2. Online Model

• Throttle Model
• Brake Model

Description

carCalibration will show you how to build a data-driven longitudinal control system by machine-learning.

Dependency

• numpy >= 1.15.4
• pandas >= 0.22.0
• scipy >= 1.2.0
• tensorflow >= 1.7.0
• keras >= 2.1.6
• python3

if you wish to use matplotlib or plotly then you should use

• matplotlib >= 2.2.2
• plotly >= 2.5.1

Usage

Input Csv

1. From Pacmod You need accel, brake, speed, steer, leftWheelSpeed, rightWheelSpeed

%time	accel([0, 1])	brake([0, 1])	speed[m/s]	steer[rad]	leftWheelSpeed[rad/s]	rightWheelSpeed[rad/s]
0	0	0.4	0	0.2	0	0
...	...	...	...	...	...	...

2. From Imu You need x, y, z direction acceleration and pitch angle

%time	x[m/s^2]	y[m/s^2]	z[m/s^2]	pitch[rad]
0	0	0	-9.8	0
....	....	....	....	....

• Prepare your csv File
• copy csv file under the data directory
• run the following command python src/main.py

Output Csv

• Result will output in the 'result/' directory
• Output csv file will be below. You will get two type csv. (Brake and Throttle)

command(Throttle or Brake)	speed[m/s]	accceleration[m/s^2]
0.0	0	0.0
....	....	....

References

https://arxiv.org/abs/1808.10134