November 2020
Please join us in welcoming Troy McMahon, who began his DIMACS postdoc on November 1. He will work with DIMACS Associate Director Lazaros Gallos on algorithms and methods for predicting the spread of diseases such as Covid-19. 
In particular, they will be looking at the effect of distance, time, and public restriction measures on geographical spreading.
Though new to DIMACS, Troy has been at Rutgers since January, working with professor Kostas Bekris and the Pracsys Group on applications of machine learning to kinodynamic planning. This work included developing learned models to represent the dynamics of complex systems. In addition to his work with Bekris and Gallos, Troy is currently teaching the undergraduate Artificial Intelligence course at Rutgers, as he also did during the 2020 spring semester.
Troy received his Ph.D. in Computer Science in 2016 from Texas A&M University under the supervision of professor Nancy Amato, where his doctoral research emphasized motion planning for constrained systems and high degree of freedom problems. After completing his Ph.D., Troy worked as a postdoc at the University of Michigan with professor Chad Jenkins on manipulation planning. As part of this work he developed the concept of wayfields to encode common actions (such as opening a drawer or picking up an object) as cost-maps overlayed onto an environment and used in combination with motion-planning algorithms to perform the action. If you want to learn more about Troy’s research in robotics links to a few of his recent papers appear at the end of this article.
Troy completed his undergraduate degree in Physics at the University of Massachusetts, and he says, "I have a strong interest in history including ancient history and US history, with a particular interest in Civil War history. I am also interested in Geology and the sciences in general."
Troy’s work at DIMACS will tap into his interdisciplinary interests, particularly his undergraduate roots in physics. He joins Gallos on a project to analyze the spreading patterns and the efficiency of quarantine measures for COVID-19. The special features of the virus and the unprecedented global response present potentially novel paths of disease transmission that have not been observed in modern times.
These paths create continuously evolving spatial and temporal patterns of observable properties, such as number of infections or virus-related deaths. Using tools from statistical physics and network science, the project seeks to understand the special features of this complex behavior and to compare these patterns in different geographical areas. The wide availability of worldwide data enables the construction of computational models that try to explain these patterns in terms of quarantine measures, timing of such measures, and environmental conditions. Such an understanding could assess the efficiency of quarantine measures implemented around the world and potentially identify the effects of temperature and humidity.
The research focuses on the scaling and correlation behavior of epidemic characteristics between different areas. The scaling analysis explores how the observed quantities change for different scales of time and space, characterized by power-law exponents. Different exponents indicate different propagation patterns which can provide information about the effectiveness of location and timing of tests and enforcement of quarantine measures. Similarly, the extent of spatial correlations and their time evolution are key indicators of spreading patterns. The analysis of the correlation lengths and correlation times can indicate how close the virus spreading process is to criticality. This is particularly important because critical systems are in general much more vulnerable to rapid spreading.
This project is supported by NSF grant DEB-2035297, with funds from the Coronavirus Aid, Relief, and Economic Security (CARES) Act. It is also part of an international collaboration with Bar-Ilan University, funded through BSF.
Learn more about Troy McMahon’s research:
Troy McMahon, Odest Chadwicke Jenkins, and Nancy Amato, Affordance Wayfields for Task and Motion Planning, Proceedings of the 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 2955-2962.
Troy McMahon, Shawna Thomas, and Nancy Amato, Sampling-based Motion Planning with Reachable Volumes for High-degree-of-freedom Manipulators, International Journal of Robotics Research 37(7), 2018.
Troy McMahon, Shawna Thomas, and Nancy Amato, Reachable volume RRT, Proceedings of the 2015 IEEE International Conference on Robotics and Automation (ICRA), pp. 2977-2984.