Click on “Download PDF” for the PDF version or on the title for the HTML version.
If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.
Robust Route Planning for Autonomous Vehicles
Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan www.asabe.org
Citation: 2022 ASABE Annual International Meeting 2200118.(doi:10.13031/aim.202200118)
Authors: Zhicheng Zhu, Yiannis Ampatzidis, Abhisesh Silwal, Panos Pardalos
Keywords: Capacitated vehicle routing, Precision agriculture, Robotics, Robust optimization, Unmanned aerial vehicles, Weed management.
Abstract. Autonomous vehicles carrying agrochemicals have been used to treat weeds in large fields. Route planning is one of the core tasks in weed management to reduce the operational cost. In this planning problem, a fleet of vehicles with limited tank capacity needs to cover each path of the field to treat different types of weeds and replenish them at the depot if needed. In the literature, the problem is usually formulated by a capacitated arc routing model, which is nonlinear and requires additional efforts for linearization. Moreover, most of the existing research assumes the weed distribution in fields is perfectly known, which is not true in practice due to the limit of current technologies and therefore leaves the robustness of the plan in question. In this research, we study this route planning problem in the presence of the weed distribution estimation error. We first provide a linear vehicle routing model to formulate the problem with deterministic weed distributions. Based on this model, we propose a robust optimization model to incorporate the uncertainty of weed distributions by using a set, call uncertainty set, that includes all possible realizations of weed distributions. The objective of the robust model becomes seeking a route feasible to all weed distribution realizations in the uncertainty set with the minimum cost. Nowadays, unmanned aerial vehicles provide a cost-effective way to estimate weed distribution. We, therefore, take advantage of this new technology by including the estimation of weed distributions from UAV in the construction of uncertainty sets. Numerical results demonstrate the need of using the robust approach and show that the proposed model leads to a better out-of-sample performance.
(Download PDF) (Export to EndNotes)