Vision-Based Control for Autonomous Vehicles

Synopsis:

This course will present an in-depth treatment on vision-based control and its application to autonomous vehicles. The maturation of synthetic vision is rapidly advancing the capability for fully-autonomous decision making to maneuver through environments with unknown obstacles. In particular, any cooperative operation will require some level of decision making for individual vehicles as well as the group to account for environmental features. This course will introduce the basics of synthetic vision and build up state-of-the-art developments in vision-based control. Techniques such as scene reconstruction and state estimation are formulated to provide feedback. Control approaches, both nonlinear and robust, are synthesized to utilize the vision-based feedback for decision making. The entire process, including path planning, is thus constructed from a series of subtasks. The authors bring their extensive experience at testing of vision-based controllers, both ground vehicles and air vehicles, through of examples and demonstrations that highlight practical issues of implementation and performance.

Key Topics:

  • Camera Characteristics
  • Scene Reconstruction
  • State Estimation
  • Robust Vision
  • Visual-Servo Control
  • Path Planning

Who Should Attend:

This course is intended for users interested in theoretical understanding of state-of-the-art concepts in vision-based control along with practical issues related to implementation. The material is appropriate for engineers with a basic knowledge of controls who want to investigate the basics of synthetic vision and its integration into autopilot structures.

Course Information:

Type of Course: Instructor-Led Short Course
Course Level: Fundamentals/Intermediate
Course Length: 2 days
AIAA CEU's available: Yes

Outline


I. Introduction

II. Cameras
a. Focal Length
b. Radial Distortion

III. Image Processing
a. Feature Point Tracking
b. Color And Texture

IV. Synthetic Vision
a. Optic Flow
b. Structure From Motion

V. Scene Reconstruction
a. Learning From Regression
b. Multiresolution Analysis
c. Updating Surface Regions

VI. State Estimation
a. Kalman Filter
b. Optic Flow Decomposition
c. Lyapunov Analysis

VII. Robust Vision
a. Camera Uncertainty
b. Calibration Uncertainty

VIII. Visual-Servo Control
a. Homography-Based Approach
b. Line-Of-Sight Progression

IX. Path Planning
a. Multi-Rate Feedback
b. Receding Horizon

X. Experimental Validation

Materials


Instructors

Rick Lind is an assistant professor in the Department of Mechanical and Aerospace Engineering at the University of Florida. He has flight tested many vision-based autopilots for UAVs and MAVs. He is the principle investigator for a MURI on vision-based control for aircraft operations in urban environments.

Andy Kurdila is the W. Martin Johnson Professor of Mechanical Engineering at Virginia Tech. He is the principle investigator for the multi-million JOUSTER program for development of autonomous ground/air vehicles. Additionally, he has over 150 publications and 4 books on control theory and dynamics.

Nick Gans is an Assistant Professor of Electrical Engineering at the University of Texas at Dallas. He has researched control of UAVs with the Air Force Research Laboratory. His research interests include nonlinear and adaptive control, with focus on vision-based control and estimation, robotics and autonomous vehicles.

 

AIAA Training Links