Machine Vision Powers Wheelchair

Researchers at Notre Dame University's Aerospace and Mechanical Engineering Department have designed a radically new approach to wheelchair navigation. For users whose disabilities cannot be met with conventional technology, the Dexterity, Vision and Control research group has designed a voice-activated, vision-based wheelchair. But what makes this system truly innovative is its use of the "teach and repeat" method commonly applied in industrial robots. Internal joint poses are calculated while a human operator physically pushes the vehicle. These poses are stored and then sequentially recalled in order to track each taught trajectory. The vision-based system or advanced prototype Computer Controlled Power Wheelchair Navigation System (CPWNS), is in testing and is being sponsored by the Department of Veterans Affairs, involving disabled veterans at the Edward Hines VA Hospital in Hines, IL

Passive vision is the technique used to combine the flow of observations with encoder readings of the wheels rotation to allow high-precision navigation. Two cameras mounted into the base of the CPWNS detect a series of landmarks or cues that are fixed to the environment in known locations, to infer the vehicle’s position and orientation. The images obtained by the cameras are manipulated and captured using a DT3162 frame grabber board from Data Translation. This board is capable of capturing 640x480 monochrome images of the video input and making them available to the host PC. When a cue is detected, an observation equation is used to incorporate the visual information to the correction of state estimates.

An earlier version of the system used the DT3152 frame grabber and the DT-Open Layers® SDK to access the image information. Recently, the hardware has been upgraded to the DT3162 frame grabber, chosen because of its high accuracy and ease-of-use. The DT3162 is accessed by way of the DT-Active Frame Grabber Control within the DT Active X control (which ships free with every DT3162), providing a much cleaner and readable code.

For more information, call (800) 525-8528.