Bilevel optimization

Bilevel Optimization is a challenging class of problems where the performance of an upper level “leader” problem is realizable only if the decision vector of a nested lower level “follower” problem is at its optimum. Alternatively, it can also be defined as an optimization problem which has another optimization problem as a constraint. Several real-life problems are hierarchical in nature and suited for modelling as bilevel optimization problem. These include (but are not limited to) transportation, policy making, supply-chain management, economics, engineering design, defence strategies, cyber-security, etc. This hierarchical nature of the problem induces several unique challenges, such as:
  • Exorbitant numbers of function evaluations are required for solving the problem since evaluation of each upper level solution requires optimization of a lower level problem. It is proven that even if both levels of the problem are linear, the resulting problem is NP-hard.
  • The leader and follower problems may be uncooperative, posing a severe difficulty for the ranking techniques used in evolutionary methods
  • Further complexity is induced when the upper/lower level problems contain multiple objectives and/or constraints.
  • Performance characterization and benchmarking of algorithms for such problems is not straightforward, since a suboptimal solution at lower level may result in a better-than-optimal solution at upper level.

Purpose of the TF

A number of studies exist in the classical literature to solve bilevel problems, but they usually involve assumptions on or knowledge of mathematical properties of the involved functions, which may severely limit their application for problems that are highly nonlinear or black-box. Evolutionary/metaheuristic and hybrid methods offer a way forward to solve such problems. The interest in use of evolutionary techniques is relatively recent, but is gaining significant traction. This task force aims to intensify efforts towards development of powerful techniques to solve difficult bilevel problems currently intractable. Some of the channels include
  • Building an online community for sharing latest developments and resources to advance the field
  • Organization of special sessions and tutorials in major events and conferences such as CEC, EMO, PPSN, SSCI, etc.
  • Organization of edited books and special issues in journals
  • Collaborative projects and papers towards solving open challenges in the field

Chairs
Ankur Sinha Ankur Sinha is an Associate Professor with the Indian Institute of Management (IIM), Ahmedabad, India. More information about his research and professional activities can be found on his webpage.

Email: asinha@iima.ac.in 		Hemant Kumar Singh Hemant Kumar Singh is an Associate Professor with the University of New South Wales (UNSW), Canberra, Australia. More information about his research and professional activities can be found on his webpage.

Email: h.singh@adfa.edu.au

Current members

(Please feel free to suggest/recommend other researchers working in the area to join)

Activities

2024/forthcoming
  • Special Session on Bilevel Optimization: Methods and Applications, IEEE Congress on Evolutionary Computation (CEC), WCCI 2024, Yokohama, Japan.
    Session chairs: H. Singh, A. Sinha, K. Deb
  • Tutorial on Evolutionary Bilevel Optimization, IEEE Congress on Evolutionary Computation (CEC), WCCI 2024, Yokohama, Japan.
    Session spekers: A. Sinha, H. Singh, K. Deb
2023 2022 2021 2020 2019
  • Special Session on Evolutionary Computation for Bilevel Optimization, IEEE Congress on Evolutionary Computation (CEC) 2019, Wellington, NZ.
    Session chairs: J. Wei, A. Sinha, K. Deb, H. Li, and M. Zhang
  • Tutorial on Evolutionary Bilevel Optimization, IEEE Congress on Evolutionary Computation (CEC) 2019, Wellington, UK.
    Organizers: A. Sinha, K. Deb
2018
  • Tutorial on Evolutionary Bilevel Optimization: An Emerging Area for Research and Application in EC, Parallel Problem Solving from Nature (PPSN) 2018, Coimbra, Portugal.
    Organizers: P. Malo, A. Sinha
  • Tutorial on Evolutionary Bilevel Optimization, IEEE Congress on Evolutionary Computation (CEC) 2018, Rio de Janerio, Brazil.
    Organizers: A. Sinha, K. Deb
  • Two bilevel optimization papers nominated in the best paper category at IEEE Congress on Evolutionary Computation (CEC) 2018, Rio de Janerio, Brazil.
  • Keynote on On Parallel Evolutionary Algorithms for Multilevel Optimization, IEEE Congress on Evolutionary Computation (CEC) 2018, Rio de Janerio, Brazil.
    Speaker: Professor Helio Barbosa, LNCC, Brazil