Bi-Objective Routing for Robotic Irrigation and Sampling in Vineyards

Abstract

Motivated by the use of robots in implementing new systems for precision irrigation, in this paper we consider a routing problem where the robot is tasked with collecting two unrelated rewards at once. While operating under a budget constraint limiting the number of locations it can visit, the robot needs to decide which locations to visit, to adjust water emitters and collect soil moisture samples to improve the inference model used to estimate soil water content. This problem can be cast as an instance of multi-objective orienteering, a computationally hard problem scarcely studied in the past. Building upon the heuristic solutions we recently developed for the single objective orienteering Problem, in this paper we develop and compare various solutions for the case involving two distinct but concurrent objective functions. The goal is to develop algorithms that are both efficient and easy to tune for a non-expert human user. Extensive simulations informed by our field experience show the effectiveness of the proposed solutions.