We conducted a 2-week long field trip to Washington state to test some of the technologies developed by CASC in orchards representative of the national apple industry,. In week 1 we focused on GPS-free row following and turning and vehicle localization. Our initial tests were performed at the Washington State University’s Sunrise Orchard in Rock Island, near Wenatchee. Sunrise was the ideal orchard for us to begin testing, as it provides a variety of row configurations – including narrow and wide and short and long rows, as well as flat and rolling terrain. We selected a relatively narrow, long set of rows located on flat terrain and worked there for three days. Despite overheating issues with the GPS-based ground-truthing system, we were able to successfully demonstrate both row following and row turning without using GPS. Perhaps one of the more important aspects was the successful demonstration of row entry using detection of the row profile prior to actual entry. We also spend significant time collecting data to further our work on GPS-free vehicle localization; we will use the data to assess the performance of the current algorithms and to design enhanced ones. Before we left Sunrise we did some quick row following tests at the more hilly parts of the orchard, not to prove that the APM works there but to collect data that will be later used to make the row following algorithms more robust to terrain inclination.

Next we moved the APM to McDougals’s Roche orchard near Moses Lake for two days of row following and turning tests in anticipation to our week 2 APM-Scout integration tests. Roche is a more challenging orchard than Sunrise in terms of row width and terrain slope; still, we were able to complete a large number of successful row following/turn runs. Turning with detected row entry works mostly well, but does occasionally fail in this orchard.  The primary cause appears to be the dense canopy on the trees at the head of the row, which prevent the APM from seeing the rest of the trees in the row and consequently estimating the row profile correctly. We also collected data for the localization work, mostly without cones, in anticipation of using scan matching or trees as landmarks.

On the way from Sunrise to Roche we stopped at two large commercial orchards to get a feeling of the future work we will need to do to adapt the APM to a broader variety of places. In these orchards, the row width is so narrow that the APM cannot traverse without touching leaves, limbs, or fruit. While we may be able to use the existing autonomy algorithms to drive in these blocks, we will certainly need to deploy them on a vehicle of a different form factor. This is actually one of the objectives of the project – to map the APM autonomy architecture to platforms that attend a variety of needs and that are able to traverse rows of different shapes and sizes.

While at Sunrise we participated in WSU’s Sunrise Research Orchard Field Day, organized by the orchard director Jay Brunner and WSU faculty. The program included talks and demonstrations on apple breeding, biological control, crop load management, robotic systems (i.e., CASC), apple replant, and orchard development.