UC Merced and UC Santa Cruz
Engineering Capstone
I submitted proposals for Engineering Capstone projects at two University of California as a means of garnering additional resources to support our autonomous construction work in Livermore, in this case specifically a compact track loader (CTL) application for moving a stockpile of material from on location to another.
​
Tasks were divided into two complementary and internally consistent scopes so that the teams could work independently. Merced was given the scope for navigation about the site and Santa Cruz the scope for filling the bucket.
​
Because multiple traversals between the source and destination stockpiles would be required and because a priori knowledge of the worksite would be limited, Merced implemented a method for estimating the cost of a traversal path and adjustment of the path for the next trip based on experience during the current and previous trips.
​
Santa Cruz implemented methods for modeling the current state of the stockpile and adjusting the direction of approach to obtain full buckets. Their method employed texture recognition to isolate the stockpile from its background and stereo vision to acquire the geometry of the stockpile. They also implemented a method to verify that the bucket was sufficiently full to warrant a trip to the destination pile. Using a rudimentary navigation solution and visual target as a destination the Santa Cruz team was able to decimate a source stockpile through as many as a dozen approach and bucket refillings without human intervention.
​
The success of these student projects made a significant contribution to Topcon’s strategy for development of autonomous equipment applications.
© 2019 Donna Kelley