Optimization-Based
Trajectory Planning
n Lecture 5
Ph.D. Candidate in Robotics
Zhejiang University
Zhepei
Wang
主讲人
Preliminaries
Global Methods: Exploration and Exploitation
Global and local methods
Local Methods: Deterministic Optimization
Global Methods: Sampling + Graph Search
Global and local methods
RRT*PRM*
JPSA*
CHOMP
DDP/iLQR
Flatness MPC/NMPC
PRM*/RRT*: https://ompl.kavrakilab.org/
A*: https://github.com/qiao/PathFinding.js/
JPS: https://github.com/KumarRobotics/jps3d
CHOMP: https://github.com/ros-planning/moveit
DDP/iLQR: https://github.com/anassinator/ilqr
MPC/NMPC: https://www.embotech.com/products/forcespro/overview/
Flatness: https://github.com/ZJU-FAST-Lab/GCOPTER
Local Methods: Deterministic Optimization
Some works try to combine both, but still inherit most disadvantages.
Global and local methods
Global Methods
Local Methods
Pros
• Global optimality
• Handling environmental complexity
• Portable
• Zero-order Information
• Local optimality
• Handling dynamics complexity
• Fast in high-dimensional space
• Promising convergence rate
Cons
• Slow in high-dimensional space
• Inconvenient to incorporate dynamics
• Poor convergence rate
• More involved
• High-order information
• Shallow local minima
