Numerati® Partners Academic Affiliate Arxiv Preprint on Drone Swarms
Numerati® Partners acknowledges the recent Arxiv research preprint Comparative Analysis of Agent-Oriented Task Assignment and Path Planning Algorithms Applied to Drone Swarms, co-authored by Rohith G. Ganesan, Samantha Kappagoda, Professor Giuseppe Loianno, PhD and Dr. David K.A. Mordecai. The paper was based on a research workstream conducted at academic affiliate RiskEcon® Lab @ Courant Institute of Mathematical Sciences NYU.
Synopsis: Autonomous drone swarms are an emerging technology with prospective applications in the field of mapping, inspection, transportation and monitoring. To complete a task, each drone in the swarm navigates a specified environment while avoiding collision with obstacles and other agents. In the specified setting, the task involves optimal coverage of an environment by the drone swarms for which global knowledge of goal states and positions, i.e., Points of Interest (PoI), are known, but not the location of obstacles. Each drone must target the PoI in the environment in order to maximize coverage. This simulation uses an agent-oriented approach to solve the problem. Different policy networks trained with reinforcement learning algorithms are evaluated based on time taken to map the computational requirements of the area. The task assignment is coupled with path planning to perform collision avoidance during navigation and compare a grid-based global planning algorithm, i.e., wavefront and a Potential Field gradient-based local planning algorithm. The Potential Field planning algorithm is evaluated across different cost functions, to propose a method to adaptively modify the velocity of a drone using the Huber loss function to perform collision avoidance and to observe its effect on the trajectory of the drones. These experiments are demonstrated in 2D and 3D simulations.
Ms. Kappagoda is Co-Managing Member of Numerati® Partners, and co-advises research at RiskEcon® Lab for Decision Metrics at Courant Institute of Mathematical Sciences. Dr. Mordecai is Co-Managing Member of Numerati® Partners, and advises research at RiskEcon® Lab @ Courant Institute. Other affiliated research papers and preprints can be accessed here: https://numeratipartnersllc.com/academic-affiliate-research/.
About RiskEcon® Lab @ Courant Institute
The mission of RiskEcon® Lab for Decision Metrics @ Courant Institute of Mathematical Sciences NYU is the development of experimental testbeds and analytics that employ high-dimensional datasets from innovative sources by applying a range of computational and analytical methods to commercial and industrial sensor networks and edge computing embedded systems, focusing primarily on research and development (R&D) of remote- and compressed- sensing, anomaly detection, forensic analytics and statistical process control. By employing applied computational statistics within the context of robust and scalable data analytic solutions, the goal is robust and reliable integration of machine learning with signal processing for measurement and control, in order to conduct research fundamental to large-scale, real-world questions in risk and liability management in the public interest.
RiskEcon® Lab for Decision Metrics was established in 2011 at Courant Institute of Mathematical Sciences, an independent division of New York University (NYU). Courant is considered to be one of the world’s leading mathematics educational and scientific research centers, and has been ranked first in research in applied mathematics. RiskEcon® Lab is the cornerstone of the Computational Economics and Algorithmic Data Analytics (CEcADA) cooperative at New York University, established concurrently in 2011.
January 14, 2021 | New York, NY
Synopsis: Autonomous drone swarms are an emerging technology with prospective applications in the field of mapping, inspection, transportation and monitoring. To complete a task, each drone in the swarm navigates a specified environment while avoiding collision with obstacles and other agents. In the specified setting, the task involves optimal coverage of an environment by the drone swarms for which global knowledge of goal states and positions, i.e., Points of Interest (PoI), are known, but not the location of obstacles. Each drone must target the PoI in the environment in order to maximize coverage. This simulation uses an agent-oriented approach to solve the problem. Different policy networks trained with reinforcement learning algorithms are evaluated based on time taken to map the computational requirements of the area. The task assignment is coupled with path planning to perform collision avoidance during navigation and compare a grid-based global planning algorithm, i.e., wavefront and a Potential Field gradient-based local planning algorithm. The Potential Field planning algorithm is evaluated across different cost functions, to propose a method to adaptively modify the velocity of a drone using the Huber loss function to perform collision avoidance and to observe its effect on the trajectory of the drones. These experiments are demonstrated in 2D and 3D simulations.
Ms. Kappagoda is Co-Managing Member of Numerati® Partners, and co-advises research at RiskEcon® Lab for Decision Metrics at Courant Institute of Mathematical Sciences. Dr. Mordecai is Co-Managing Member of Numerati® Partners, and advises research at RiskEcon® Lab @ Courant Institute. Other affiliated research papers and preprints can be accessed here: https://numeratipartnersllc.com/academic-affiliate-research/.
About RiskEcon® Lab @ Courant Institute
The mission of RiskEcon® Lab for Decision Metrics @ Courant Institute of Mathematical Sciences NYU is the development of experimental testbeds and analytics that employ high-dimensional datasets from innovative sources by applying a range of computational and analytical methods to commercial and industrial sensor networks and edge computing embedded systems, focusing primarily on research and development (R&D) of remote- and compressed- sensing, anomaly detection, forensic analytics and statistical process control. By employing applied computational statistics within the context of robust and scalable data analytic solutions, the goal is robust and reliable integration of machine learning with signal processing for measurement and control, in order to conduct research fundamental to large-scale, real-world questions in risk and liability management in the public interest.
RiskEcon® Lab for Decision Metrics was established in 2011 at Courant Institute of Mathematical Sciences, an independent division of New York University (NYU). Courant is considered to be one of the world’s leading mathematics educational and scientific research centers, and has been ranked first in research in applied mathematics. RiskEcon® Lab is the cornerstone of the Computational Economics and Algorithmic Data Analytics (CEcADA) cooperative at New York University, established concurrently in 2011.
January 14, 2021 | New York, NY
