Planning for safe and efficient operations of an aerial-platform based semiautonomous pre-harvest crop inspection system

Project description

We need to generate safe and efficient multi-vehicle inspection task plans by taking into account (1) human cognitive load constraints, (2) farm environment constraints, and (3) user-specified information about risky areas, such as areas with the possible presence of fecal matter. We need to generate energy-efficient coverage trajectories for a given task by taking into account wind conditions. For these purposes, we define a multi-robot inspection problem, where the task of a cooperative team of quadrotors and a human operator is to efficiently and accurately inspect a field of an arbitrary shape for the signs of animal intrusion and the presence of fecal matter under environmental uncertainties. We anticipate the problem to be, in large part, formulated as a multi-objective optimization problem. The critical part of the algorithm will be to partition the field into arbitrarily shaped cells (i.e., cells with curved boundaries) based not only on the geometry of the field and distribution of static obstacles but also on the knowledge of the high-risk areas. The computation of efficient scanning patterns for cells will consist of general curves in order to achieve higher scanning efficiency and limit the need for sharp turns. Since there are many possible combinations of scanning patterns across the cells, an exciting possibility will be to develop an optimization algorithm that computes an approximation of a theoretically optimal scanning pattern over the entire field such that the traversal overhead between the cells is minimized. We will develop a system-level energy consumption model that will estimate the power consumption of a quadrotor in a strong wind field. The cost function used during the computation of scanning patterns for individual cells will incorporate the direction and magnitude of the wind field.

Outcomes

The successful completion of this project will yield the following outcomes:
a) Planning algorithm for decomposition of the field into cells of arbitrary shapes, computation of efficient scanning patterns for cells, and sequential allocation of quadrotors for visiting the cells
b) Preemptive task planning for handling task execution contingencies and matching the available cognitive capacity of a human operator
c) Utilizing user-provided information about high-risk areas to increase scanning reliability and incorporating the effects of strong winds into inspection planning to increase scanning efficiency
d) Research publications in international conferences and journals

Essential capabilities

Algorithmic thinking and demonstrable programming skills in Matlab and Python

Desireable capabilities

Ability to implement and understand typical variants of A* and RRT algorithms

Expected qualifications (Course/Degrees etc.)

B.Tech. in any relevant field, or bachelor’s degree in mathematics or statistics, or Master’s degree in technology (MS or M.Tech.) in any relevant engineering field