Using Machine Vision Perception to Control Automated Vehicle Maneuvering

Automated Vehicles (AVs) for aerial, ground, water or space applications must have the capability to establish their location relative to a destination in order to determine the necessary direction of travel.  Engineers often employ technologies such as celestial, inertial, or electronic navigation to monitor and control moving AVs from one place to another.  However, often AVs must also follow uncharted paths or avoid obstacles, change speed in response to local conditions, maintain very precise positioning, or operate without external navigation aids.  Additional sensors and systems are needed to operate under these conditions.  Several technologies are available as onboard sensors, but their size, weight, power, or cost are often prohibitive.  The favorable size, weight, power, and cost of cameras for machine vision perception makes this technology attractive for many AV applications.  This lecture focuses on using machine vision perception to control AV maneuvering.

The lecture will include several applications, including Ranger: a ground-facing camera-based localization system for AVs.  R&D magazine recognized Ranger as being among the 100 most significant innovations in 2015.  The following year, Ranger was recognized as the best paper at the AESS/IoN Position Location and Navigation Symposium (PLANS).

Other examples that will be discussed include:

• An unmanned aerial vehicle to inspect the interior of a damaged nuclear reactor
• An unmanned ground vehicle that follows humans using hand signals
• Using machine vision to recognize humans walking, riding bicycles, etc.
• Classifying obstacles (e.g. boulders, trees, shrubs, grass, ditches, etc.) with machine vision
• Using machine vision to inspect roadways
• Autonomous parking of tractor trailers
• Forklift operations
• Agricultural applications
• Other off-road applications