Abstract. A two-unit experimental sweet cherry harvester was modified with a goal to improve its harvest performance by changing each unitâ€™s rapid impact fruit removal mechanism to a continuous impact (shaking) mechanism. During harvest, an end effector (rapid impactor) was placed at selected actuation points on fruiting scaffold branches to transfer vibrational energy for fruit release. Removed fruit was caught by a catching system and was collected into a lug or bin placed on a platform. Remote control capability was incorporated into each harvester unit to improve its operability. This article reports details on these added features and results from field evaluation studies conducted in 2012 and 2013 harvest seasons in Washington State. Harvest performance was evaluated in terms of harvesting rate and harvest efficiency. Harvesting rate consisted of two parts: tree coverage rate (number of trees harvested per unit time) and fruit removal rate (weight of fruit removed per unit time). Harvest efficiency included fruit removal efficiency (weight of removed fruit per weight of total fruit on tree) and collection efficiency (weight of machine-caught fruit per weight of removed fruit). In the first of two tests conducted in 2012, using only the left half of the harvester to harvest the left half of trees, average fruit removal efficiency of 83% was achieved and of those removed, over 85% was collected. Tree coverage rate was on average 19 and 15 trees/h, respectively, for 10 and 15 s shaking per actuation point with fruit removal rate of 7.4 kg/min. In the second test conducted in 2012, tree coverage rate averaging 30 trees/h was achieved. It was observed in 2012 tests that a variable shaking duration would be necessary to not shake branches after most of the cherries are detached. The 2013 tests used actuation point-specific shaking periods on Ethephon-treated trees, resulting in an average tree coverage rate of 36 trees/h, removal efficiency of 87%, and collection efficiency of 79%. Harvest-induced fruit damage in 2013 tests ranged from 0% to 20% with an average of 10%, which was better or comparable to previously reported values. Use of remote control, compared to control from the operatorâ€™s seat, increased operatorâ€™s visibility of target branches up to two times and contributed to 16% less time with 9% more accuracy in positioning the continuous impactor on target branches. These results show promise for practical applicability of this harvest system in commercial operation.