Energy Efficient Planning for Repetitive Heterogeneous Tasks in Precision Agriculture

TL;DR

The paper addresses energy-efficient planning for repetitive heterogeneous tasks in precision agriculture by casting robotic weed removal as RHTP under an Observe-First-Manipulate-Later constraint. It introduces a Probabilistic Target Reachability Map and a Task Space Partition to transform reachability and clustering into a region-based MINLP solved with Branch-and-Bound, thereby minimizing the long-run energy cost per cycle. Key contributions include the PTRM, partition-based region formulation, and a MINLP/SET-COVER–like approach that yields significant improvements in path length, stops, energy, and replans, especially at higher target densities. The approach demonstrates practical impact for energy-conscious field robotics by leveraging known spatial distributions and STM to enable efficient repetitive tasks in precision agriculture.

Abstract

Robotic weed removal in precision agriculture introduces a repetitive heterogeneous task planning (RHTP) challenge for a mobile manipulator. RHTP has two unique characteristics: 1) an observe-first-and-manipulate-later (OFML) temporal constraint that forces a unique ordering of two different tasks for each target and 2) energy savings from efficient task collocation to minimize unnecessary movements. RHTP can be framed as a stochastic renewal process. According to the Renewal Reward Theorem, the expected energy usage per task cycle is the long-run average. Traditional task and motion planning focuses on feasibility rather than optimality due to the unknown object and obstacle position prior to execution. However, the known target/obstacle distribution in precision agriculture allows minimizing the expected energy usage. For each instance in this renewal process, we first compute task space partition, a novel data structure that computes all possibilities of task multiplexing and its probabilities with robot reachability. Then we propose a region-based set-coverage problem to formulate the RHTP as a mixed-integer nonlinear programming. We have implemented and solved RHTP using Branch-and-Bound solver. Compared to a baseline in simulations based on real field data, the results suggest a significant improvement in path length, number of robot stops, overall energy usage, and number of replans.

Figures (6)

• Figure 1: A weed removal robot operating in a cotton field, where the target weeds are highlighted as colored regions. Weed regions sharing the same color belong to the same cluster. The RHTP algorithm optimizes observation and weed flaming task.
• Figure 2: (a) A running example with 4 targets and the TROIs. (b) PTRM (2D at x-y plane). These probabilities are superimposed on one coordinate system.
• Figure 3: PTRM partition. Left: a topological visualization of task space partition. Right: partition's task success probabilities matrix. These probabilities are based on location and not normalized against either $i$ or $j$.
• Figure 4: Sample scene configurations with different target densities from early/late grow stage in cotton field and test bed.
• Figure 5: Performance comparison between the Naive-CAPM and the RHTP algorithms.

Theorems & Definitions (3)

• Definition 1
• Definition 2
• Definition 3