Improving Quality of Fruits Picked in Orchards Using Mahalanobis Distance Based Model

Abstract

A challenge in fruit orchards is the collection of fruits lying on the floor. They need manual collection which is time-consuming and labor-intensive. Automated collection is desired in such a scenario. A predefined collection path is required for automated collection using robots. A novel path planning methodology is introduced to perform nonuniform coverage, and paths are evaluated with the information available about fruit distribution probability in the environment. A fruit probability map was developed describing the potential of fruits at every location. A mathematical model defining the shortest path to a fruit distribution using Mahalanobis distance is evaluated. This path is compared with the paths used by a farmer for the collection of fruits. The amount of time spent by fruits lying in the ground has to be minimized. An even bigger challenge is the decay of fruits lying in the ground. Though optimality of the path is not guaranteed with respect to fruit distribution, usage of Mahalanobis distance model has significantly reduced the time spent by fruits lying in the floor. The presented results are promising for the usage of automated robots to collect fruits. This will lead to a reduction in the demand for manual labor. Extending the model with elevation data could further improve the results by increasing energy efficiency and decreasing time consumed.