1. Dive into Self-Driving Cars
technologies using ROS and
Open Source software
Humanoid Robotics, Artificial Intelligence
and Automation
21 May 2020
TheRobotAcademy.com
3. 3
INDEX
Self Driving Cars in context
Decision making 1:
State Machines
Decision making 2:
Reinforcement Learning
Decision making 3:
Relational Learning
1
2
3
4
5. 5
The 5 levels of Autonomous Vehicles
Self Driving Cars in context
Level 1 - No Automation Driver keeps full control: ordinary cruise control
6. 6
Level 1 - No Automation
Level 2 - Partial Automation
Assist in controlling speed and steering:
stop-and-go traffic & centering in the lane
7. 7
Level 1 - No Automation
Level 2 - Partial Automation
Level 3 - Conditional Automation
Self-driving under ideal conditions and with limitations:
Off ideal -> driver is required to take over
8. 8
Level 1 - No Automation
Level 2 - Partial Automation
Level 3 - Conditional Automation
Level 4 - High Automation
Drive themselves without human interactions:
restricted to known use cases
9. 9
Level 1 - No Automation Driver keeps full control: ordinary cruise control
Level 2 - Partial Automation
Assist in controlling speed and steering:
stop-and-go traffic & centering in the lane
Level 3 - Conditional Automation
Self-driving under ideal conditions and with limitations:
Off ideal -> driver is required to take over
Level 4 - High Automation
Drive themselves without human interactions:
restricted to known use cases
Level 5 - Full Automation
True driverless cars:
No need for a steering wheel and pedals
The 5 levels of Autonomous Vehicles
11. 11
Stimulus- Response
State machine
(no learning)
rules are preprogrammed (hard coded)
Given a concrete set of values of state variables,
always produce the same action
Level 1: State Machines
Decision making I
12. 12
Level 1: State Machine case study
Car switch lights
● (trigger) If state is OFF, (action) then ON when
○ (detection) Natural lighting is reduced <=
■ (scenario) dusk till dawn || tunnel || fog
● (trigger) If state is ON, (action) then OFF when
○ (detection) There’s daylight <=
■ (scenario) daylight
○ (detection) Intense light source <=
■ (scenario) Lights from another car traveling in the opposite direction