4. Types of Suspension
• Passive Suspension
Standard Spring and Damper
• Active
Force Generator Replaces Damper
• Semi-Active
Controllable Dampers
5.
6. Motivation for Learning Control
• Difficult System To Model
• Non-linear
• Interaction with Engine Modes
• Noise Disturbance Input
Measurement Noise
• Uncertainty In System Parameters
• System Identification is an Approximation
8. Pseudo Code
Repeat
Generate Actions
Execute The Actions On The
Environment
Receive Performance Evaluation For This
Action
If "Favourable" Then
Promote Action In Future
Situations
Until Convergence Criterion Satisfied.
9. CONTINUOUS ACTION
REINFORCEMENT LEARNING
AUTOMATA ( CARLA )
• Related To Discrete Learning Automata
• Each CARLA Operates On A Single Action
Variable
• Continuous Parameter Range
• Reward/Inaction Based
• Suitable For Stochastic Systems
• Distributed Control
• Simple Implementation