Abstract
This work addresses task planning under uncertainty for precision agriculture applications whereby task costs are uncertain and the gain of completing a task is proportional to resource consumption (such as water consumption in precision irrigation). The goal is to complete all tasks while prioritizing those that are more urgent, and subject to diverse budget thresholds and stochastic costs for tasks. To describe agriculture-related environments that incorporate stochastic costs to complete tasks, a new Stochastic-Vertex-Cost Aisle Graph (SAG) is introduced. Then, a task allocation algorithm, termed Next-Best-Action Planning (NBA-P), is proposed. NBA-P utilizes the underlying structure enabled by SAG, and tackles the task planning problem by simultaneously determining the optimal tasks to perform and an optimal time to exit (i.e. return to a base station), at run-time. The proposed approach is tested with both simulated data and real-world experimental datasets collected in a commercial vineyard, in both single- and multi-robot scenarios. In all cases, NBA-P outperforms other evaluated methods in terms of return per visited vertex, wasted resources resulting from aborted tasks (i.e. when a budget threshold is exceeded), and total visited vertices.
Highlights
A UTONOMOUS agricultural mobile robots become increasingly more capable for persistent missions like monitoring crop health [1] and sampling specimens [2] across extended spatio-temporal scales to enhance efficiency and productivity in precision agriculture [3]
This paper introduces a new stochastic task allocation approach, termed Next-Best-Action Planning (NBA-P), for task planning under uncertainty in precision agriculture
NBA-P is compared against lawnmower planner [28], which is often seen in agriculture-related applications [29, 30]
Summary
A UTONOMOUS agricultural mobile robots become increasingly more capable for persistent missions like monitoring crop health [1] and sampling specimens [2] across extended spatio-temporal scales to enhance efficiency and productivity in precision agriculture [3]. As the budget is being depleted, the robot needs to periodically return to a base station (e.g., to drop collected samples and/or recharge) Addressing these two challenges simultaneously poses a two-layer intertwined decision making under uncertainty problem: How to perform optimal sampling given an approximate prior map, and how to decide an optimal stopping time (i.e. to return to base) to avoid exceeding a given task capacity? This paper introduces a new stochastic task allocation algorithm to balance optimal sampling and optimal stopping when task costs are uncertain. The robot prioritizes visiting adjacent locations if they jointly yield higher gains than isolated high-gain locations, and provided that any budget constraints are not violated [16, 17] This strategy can be insufficient for missions where some tasks are more urgent than others. NBAP achieves higher efficiency than naive lawnmower, informed lawnmower, and series Greedy Partial Row planners [28,29,30] in terms of more return per visited vertices, less resources wasted because of aborted tasks, and less total visited vertices
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call