Abstract: I will spotlight our recent result in visual-perceptive navigation of mobile robots in outdoor field environments. Small mobile robots hold a lot of potential for many outdoor applications, including agriculture, defense, disaster recovery, and remote area exploration. However, prevalent waypoint-based path programming methods are ineffective and cumbersome. This is a major barrier for deployment of field robots because users have to plan paths around all possible obstacles, including rocks, trees, holes, etc.
I will describe our recent work in visual-perceptive navigation, starting first with visual-navigation in semi-structured agricultural fields, and generalizing to traversability estimation in forests, grasslands, and near beaches. I will also outline some open challenges in field robotics, ranging from perceptive navigation to Edge computation, and look forward to a healthy discussion of what the CS community can do to make field robots more practical and ubiquitous. Finally, I will describe the facilities at the Illinois Autonomous Farm, and describe how you can get involved.