Visual Odometry Using Pixel Processor Arrays for Unmanned Aerial Systems in GPS Denied Environments.

Area of Science:

• Robotics and Autonomous Systems
• Computer Vision
• Sensor Fusion

Background:

• Autonomous systems struggle with navigation in Global Positioning System (GPS) or Global Navigation Satellite System (GNSS) denied environments.
• Operating in challenging locations like indoors, urban canyons, or natural canyons is often impossible for current systems.
• Existing navigation methods are heavily reliant on consistent GNSS signal availability.

Purpose of the Study:

• To demonstrate the use of a Pixel Processor Array (PPA) vision chip for real-time navigation and control of aerial robots.
• To address the limitations of GNSS-impaired environments for autonomous systems.
• To showcase the SCAMP vision chip's capability in combining perception and computation on a single device.

Main Methods:

• Utilized a SCAMP PPA vision chip, integrating processing elements with light capture, processing, and storage.
• Implemented concurrent visual odometry and target identification directly on the PPA device.
• Conducted outdoor multirotor test flights to evaluate performance against baseline GPS.

Main Results:

• Achieved a combined frequency of approximately 400 Hz for visual odometry and target identification on the PPA.
• Demonstrated successful real-time navigation and control in GNSS-challenging conditions.
• Showcased the SCAMP PPA's High Dynamic Range (HDR) capability for varied lighting conditions.

Conclusions:

• The SCAMP PPA vision chip effectively aids aerial robot navigation in GNSS-denied environments.
• Its ability to perform on-sensor processing significantly reduces bandwidth requirements and enables low-power operation.
• The PPA's HDR and adaptive rate capabilities make it suitable for robust outdoor flight in challenging conditions.