Current progress and boundaries
Harvesting robots have been prototyped for specific crops and structured environments, but reliable deployment across crops and seasons remains challenging.
Why it remains difficult
Robots must combine 3D perception, fine manipulation, and path planning. Moving too slowly affects economic viability, while applying the wrong force can damage fruit.
Paths worth exploring
Break harvesting into localization, ripeness assessment, grasping, and quality checks. Establish standardized tests for one variety before gradually expanding the range of conditions.
How to evaluate success
Measure success per fruit, damage rate, fruits harvested per hour, time to adapt to a new field, and maintenance costs; compare results with a human baseline.