INFORMS Open Forum

In many areas of science and engineering inverse problems arise. These problems consider a set of field observations and aim to derive the causal factors that generated the data. Inverse problems are difficult to tackle because they are underdetermined, i.e., there is often a, possibly infinite, set of causal factors that are consistent with the field data. These solutions are then used to tackle control problems. This project aims to characterize the set of solutions and to quantify their impact on the solution set of the control problems. This ultimately enables the design of robust causal inference and robust counterfactual analysis techniques. 

This project  develops compute-efficient optimization techniques to tackle these problems. The project can cover research areas such as bilevel optimization, fixed-point problems, game theory, traffic assignment, travel demand estimation. 

The project is motivated by high-dimensional inverse problems that arise in the field of urban mobility and transportation. The inverse problems consider traffic field data that describes congestion patterns and aim to  derive the travel demand that generated these congestion patterns. Control problems include congestion pricing and traffic management problems. 

Research Team. 

The researcher will be co-advised by Prof. Dario Paccagnan (Department of Computing, Imperial College London) and Prof. Carolina Osorio  (Scale AI Research Chair, Department of Decision Sciences, HEC Montreal; Google Research). Academic affiliations can be at either Imperial College or at HEC Montreal with potential affiliations to various research centers (e.g., CIRRELT, GERAD) and  numerous career opportunities within the Scale AI and the IVADO ecosystems. Details on the PhD program at Imperial College can be found at https://www.imperial.ac.uk/computing/prospective-students/phd/. Click or tap if you trust this link." data-safelink="true" data-linkindex="0" data-ogsc="">link. Similarly, for HEC Montreal, see link.