Scribd
Upload
47 views

08 Trajectoryoptimization1 Slides

1. Trajectory planning involves generating a trajectory as a function of time for a robot to navigate along, as opposed to just a path with waypoints but no timing information. 2. There are two main approaches to trajectory planning - direct planning that optimizes the path and trajectory simultaneously, and decoupled planning that first plans a path then generates the trajectory. 3. Trajectory optimization involves specifying constraints like boundary conditions and continuity conditions between trajectory segments to generate smooth trajectories for robots to follow between waypoints.

Uploaded by

Bavantha Udugama
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
47 views

08 Trajectoryoptimization1 Slides

1. Trajectory planning involves generating a trajectory as a function of time for a robot to navigate along, as opposed to just a path with waypoints but no timing information. 2. There are two main approaches to trajectory planning - direct planning that optimizes the path and trajectory simultaneously, and decoupled planning that first plans a path then generates the trajectory. 3. Trajectory optimization involves specifying constraints like boundary conditions and continuity conditions between trajectory segments to generate smooth trajectories for robots to follow between waypoints.

Uploaded by

Bavantha Udugama
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 15

16.

485: VNAV - Visual Navigation for


Autonomous Vehicles
Lecture 8: Trajectory Optimization
Luca Carlone
Planning vs. Control

desired
trajectory control robot’s
inputs state
+ Controller Robot
-

Sensors
Planning vs. Control

desired
trajectory control robot’s
inputs state
Planning + Controller Robot
-
initial
and
goal
state
Sensors
B
obstacle

obstacle obstacle

obstacle

A
Trajectory Planning vs. Path Planning
• Path vs. trajectory:
• A path usually contains a sequence of waypoints, without time label
or information about velocity or higher order of derivatives.
• A trajectory defines the path the robot should navigate as a function
of time.
• Path planning:
• “slow” systems
• Many approaches: graph-based algorithms (PRM, A* and variants),
sampling-based methods (RRT, RRT* and variants)
• Trajectory planning:
• Key for high-speed maneuvers on drones
Path / Motion Planning
• Open-source libraries
• The Open Motion Planning Library (OMPL)
• http://ompl.kavrakilab.org/
• Motion Strategy Library (MSL)
• http://msl.cs.uiuc.edu/msl/
• RRT* Library
• https://svn.csail.mit.edu/rrtstar/
• Sampling Based Planning Library
• https://svn.csail.mit.edu/smp
• References:
• Howie Choset , Kevin Lynch et al, “ Principles of Robot Motion” MIT press, 2005.
• Steven Lavalle, “ Planning Algorithms”, Cambridge University Press, 2006.
Path / Motion Planning Example

• RRT*:
Rapidly
exploring
Random
Trees
Path / Motion Planning

COMBINATORIAL PLANNING SAMPLING BASED PLANNING

❑ (resolution) complete
▪ they find the solution if ❑ Probabilistic Complete
exists, otherwise they ▪ The probability of finding
report failure
a solution if it exists tends
❑ No need for collision
checking, since is to one, otherwise it may
explicitly defined run forever.
❑ Expensive computation,
impractical for higher ❑ More practical
dimensional systems ❑ Narrow passage challenge
▪ Exploding number of cells ❑ The output path is usually
❑ The output path is usually
jagged and needs jagged and needs
smoothing smoothing
Trajectory Planning: Two Approaches
• Direct trajectory planning: plan the
path and optimize trajectory
simultaneously
• Typically slow for online planning
and fast dynamics (difficult to A
account for obstacles)
• Fast heuristics (next slide): B
motion primitives

• Decoupled trajectory planning: first


path planning then trajectory
generation
• Fast computation (only partially A
accounts for obstacles)
• Polynomial trajectory optimization B
Motion Primitives
• Input:
• A library of motion primitives: each
motion primitive is generated by
forward simulation of the system for a
given control sequence
• The current robot state
• Output: a choice of motion primitive that
minimizes a cost function while avoiding
obstacles.

• The final trajectory is usually not smooth


• More complex instantiations:
https://www.cs.cmu.edu/~maxim/files/
tutorials/robschooltutorial_oct10.pdf
Output: 1->1>4->1->1->6
Motion Primitives For Aggressive Flight

2017
Motion Primitives For Aggressive Flight

2019
Decoupled Trajectory Planning
Intermediate
waypoints
desired
trajectory control robot’s
Path Trajectory inputs state
planning optimization
+ Controller Robot
-

Trajectory Planning
Sensors

B
Trajectory Optimization

2013
Trajectory Optimization

2016-2017
Trajectory Optimization
x
x1(t1) = x2(t1) = x̄1
x· 1(t1) = x· 2(t1) x2(t2) = x3(t2) = x̄2
x··1(t1) = x··2(t1) x· 2(t2) = x· 3(t2) x3(t3) = x̄3
x··2(t2) = x··3(t2) x· 3(t3) = v̄3

x1(t0) = x̄0
x· 1(t0) = v̄0

t0 t1 t2 t3
t
x1(t) x2(t) x3(t)

You might also like