Distributing Potential Games on Graphs Part II. Learning with application to platoon matching
| Autor: | Mohamed I. El-Hawwary, Jonas Mårtensson |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
Computer Science::Computer Science and Game Theory
0209 industrial biotechnology Theoretical computer science Matching (graph theory) Basis (linear algebra) Computer science 020208 electrical & electronic engineering 02 engineering and technology Outcome (game theory) symbols.namesake 020901 industrial engineering & automation Control and Systems Engineering Nash equilibrium Convergence (routing) 0202 electrical engineering electronic engineering information engineering symbols Platoon State (computer science) Potential game |
| Zdroj: | IFAC-PapersOnLine. 53:6703-6708 |
| ISSN: | 2405-8963 |
| DOI: | 10.1016/j.ifacol.2020.12.094 |
| Popis: | In part I of the paper the problem of distributing potential games over undirected graphs was formulated. A restricted information potential game was designed using state-based formulation. Here, learning Nash equilibria for this game is studied. An algorithm is developed with mainly two phases, an estimation phase and a learning phase. The setting allows for available learning methods of the full information game to be directly incorporated in the learning phase. The result matches the outcome (i.e. converges to the same equilibria) of the full information game. In addition, the design takes into account considerations of convergence time, and synchrony of actions update. The developed distributed game and learning algorithm are used to solve a platoon matching problem for heavy duty vehicles. This serves two objectives. First, it provides a motivation for the presented gaming results. Second, the problem addressed can facilitate platoon matching where it provides a basis for an on-the-go strategy. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|