Evolving Fuzzy Models and Transportation Applications
Radu-Emil Precup, Department of Automation and Applied Informatics, Politehnica University of Timisoara, Romania
As specified in the classic papers on evolving fuzzy systems due to Angelov and his co-authors back in 2001 and 2002, a specific feature of evolving fuzzy models in their either Mamdani or (Takagi-Sugeno or Takagi-Sugeno-Kang fuzzy) forms is the continuous online learning of the rule base. These fuzzy models are derived by online identification algorithms. The online identification algorithms continuously evolve the parameters in different subsystems of the fuzzy models, which are built online by adding new or removing old so-called "local" models. The local models are placed in the rule consequents, and this process is referred to as the adding mechanism.
The well-recognized classification of online identification algorithms by Dovžan, Logar, and Škrjanc in 2015 highlights three representative families of online identification algorithms, (i), (ii), and (iii): (i) adaptive algorithms – which start from the initial Takagi-Sugeno-Kang fuzzy model structure given by other algorithms (i.e., (ii) incremental algorithms – which implement adding mechanisms; (iii) evolving algorithms – these are the most advanced and fresh ones, as they implement, besides the adding mechanism, also removing and a part of them merging and splitting mechanisms.
The operating mechanisms of online identification algorithms briefly described above give valid explanations on the rather large area of nonlinear systems whose behavior is characterized in order to ensure both accurate modeling for simulation and model-based fuzzy control design. The scope of the development of these models is the model-based and data-driven model-free design and tuning of fuzzy controllers by the Process Control Group of the Politehnica University of Timisoara, Romania.
This presentation focuses on some of the results obtained by the Process Control Group in applications of evolving fuzzy models. The presentation covers some representative and recent applications implemented in the group's laboratories, with real-world validation through experiments and detailed simulations. The results exemplified in this presentation concern several transportation applications, including the speed control of Connected Autonomous Electric Buses, taking into account the presence of human-driven vehicles, the anti-lock braking systems, and several servo systems.
This work was supported by a grant of the Ministry of Research, Innovation and Digitization, CNCS/CCCDI - UEFISCDI, project number ERANET-ENUAC-e-MATS, within PNCDI IV.
Keywords—evolving fuzzy models, incremental algorithms, Takagi-Sugeno-Kang fuzzy models, transportation applications.