Set Point Stabilisation
maintain system at specific state
Thermostat
Geometric Planning
continous
time and dynamics not considered
Kinodynamic Planning
like geometric planning
time and dynamics considered
State Estimation
Estimate 𝒙𝑘 based on dynamics and measurements
Discrete Planning
finite states
Model-Predictive Control
Building a model
Choosing a cost function
Solving an optimization problem
Iterative Learning Control
simple, but Reference tracking only
used when input and output trajectory given
Reinforcement Learning
reward function
Advantages
RL can solve almost any control/motion task
One method can solve many different tasks
Disadvantages
Hours/days of system interaction and millions of trials required!
Not real-world applicable!
Behavior Cloning
Bayesian optimization
optimierung von hyperparametern.
Nutzung von GP
globale optimisation
Zuletzt geändertvor 4 Monaten
