FACILITY LOCATION

2013/08 - 2017/04

With Professor Yanfeng Ouyang

University of Illinois

CORRELATED FACILITY DISRUPTION

  • Designed an overarching model framework for facility location design under correlated facility disruptions
  • Introduced the idea of supporting station structure to decompose any type of facility correlations
  • Proposed a mixed-integer model and Lagrangian relaxation based algorithms to solve the RFL problem

RELIABLE SENSOR DEPLOYMENT

  • Optimized sensor deployment locations to maximize system-wide positioning/surveillance benefits
  • Incorporated the risk of site-dependent probabilistic sensor disruptions into the optimization framework
  • Developed a compact mixed-integer programming model and designed several customized algorithms

RELIABLE NETWORK DISTRICTING

  • Designed mixed-integer models for reliable network districting problems with multiple practical criteria
  • Developed both customized heuristic algorithms and column-generation based algorithms
  • Used by CSX Transportation Inc to solve their large-scale districting applications (e.g., call center design)

RAILROAD PLANNING

2013/10 - 2016/10

With Professor Yanfeng Ouyang & Kamalesh Somani

University of Illinois & CSX Transportation

LOCOMOTIVE INFRASTRUCTURE

  • Developed large-scale mixed-integer mathematical models and decomposition-based solution algorithms
  • Numerous field studies demonstrated significant superiority of the model solutions over the current practice
  • Used proactively by CSX Transportation Inc and have realized annual savings of more than $10 millions

NETWORK TRAVERSAL SCHEDULING

  • Designed mathematical programs and heuristic algorithms to schedule periodic traversal of network vertices
  • Developed graph packages with implementations of multiple data structures and graph algorithms (e,g., BFS/DFS, Dijkstra’s, Floyd-Warshall, Ford-Fulkerson)
  • Used by CSX Transportation Inc to solve their grinder scheduling problems

MACHINE LEARNING

2015/09 - 2016/12

With Professor Yanfeng Ouyang & Zhaodong Wang

University of Illinois

FAILURE PREDICTION

  • Applied machine learning techniques (e.g., SVM, kernel) to learn an traffic/railroad incident prediction model
  • Used customized simulation-based over-sampling technique to handle the issue of data unbalance
  • Implemented the cross-validation approach in Matlab to evaluate the model in terms of precision and recall

SIGNAL TIMING

2011/11 - 2014/05

With Professor Hai Jiang & Chiwei Yan

Tsinghua University

PHASE SWAP STRATEGY

  • Proposed an operational paradigm for the phase swap sorting strategy to increase capacity of intersections
  • Built a mixed-integer model to formulate the paradigm with both geometric and signal timings decisions
  • Implemented the paradigm into simulation programs to visualize and validate the proposed model solutions

UNCONVENTIONAL INTERSECTION DESIGN

  • Proposed a new unconventional intersection design to increase the capacity of isolated intersections
  • Built a mixed-integer model to formulate the paradigm with both geometric and signal timings decisions
  • Conducted numerical experiments to quantify the benefit of the proposed design against the conventional design