Achieving a robust position and orientation estimate is crucial for intuitive interaction with autonomous systems, especially through augmented reality interfaces. However, available passive localization methods in GPS-denied environments do not suffice. This project loosely coupled inertial and visual sensors by modifying the monocular ORB SLAM algorithm. Data collected from LIDAR and motion capture was used to evaluate the realized system. ORB SLAM code was analyzed and performance profiled for real-time implementation. SLAM scale uncertainty was corrected with inertial data, and scale drift correction was attempted by modifying an internally-optimized motion model. A more accurate position estimate was achieved, and additional work can improve precision, robustness, and execution speed.

• This report represents the work of one or more WPI undergraduate students submitted to the faculty as evidence of completion of a degree requirement. WPI routinely publishes these reports on its website without editorial or peer review.

Creator

• Bove, Christopher Peter
• Wald, Alexander William

贡献者

• LaPenta, Jason
• Donahue, Mark

Publisher

• Worcester Polytechnic Institute

Identifier

• E-project-110416-101803

Advisor

• Michalson, William R.

Year

• 2016

Center

• MIT Lincoln Laboratory-Lexington, Massachusetts Project Center - MQP

Sponsor

• MIT Lincoln Laboratory

Date created

• 2016-11-04

Resource type

• Major Qualifying Project

Major

• Computer Science
• Robotics Engineering

Rights statement

• In Copyright

最新修改

• 2023-09-28