Paper in at - Automatisierungstechnik
- publications
Cooperative control plays an important role in running dynamical multi-agent systems (MAS). However, the process of joint decision making inherently requires the agents to exchange data within the MAS. In many cases, the information to be shared is considered senstive such that the agents would like to keep it private.
Our paper addresses these privacy issues within the framework of distributed optimization. The method is based on solving a general consensus type optimization problem via the alternating direction method of multipliers (ADMM) which is reformulated such that it can be implemented securely using a leveled homomorphic encryption scheme. Our approach stands out for supporting the implementation of fine-grained access rights effectively respecting ownership of data and protecting the individual agents' private information. The main ingredient here is a careful co-design of the optimization algorithm and the strategy to distribute the involved cryptographic keys. The implementation hinges on the technique of homomorphic key switching. We tested our algorithm in a privacy-preserved robot formation control scheme.
For further details, please contact Philipp, Janis, or Nils and consider looking at the paper:
P. Binfet, J. Adamek, N. Schlüter and M. Schulze Darup. Towards privacy-preserving cooperative control via encrypted distributed optimization, at - Automatisierungstechnik, vol. 71, no. 9, pp. 736-747, 2023. DOI: 10.1515/auto-2023-0082