Effective energy and mobility management benefits from multi-agent simulations (MAS) to model complex interactions among various agent types. Selecting the optimal MAS platform to implement and simulate these interactions is vital for achieving accurate results, scalability to realistic problem sizes, and efficient computational performance. This paper investigates the energy and mobility domain and identifies key parameters such as the number and complexity of agents, parallel computing power, CPU requirements etc., for developing MAS in the context of this domain. It then presents a comprehensive evaluation of various MAS development platforms. Using a multi-level selection and elimination approach, we narrowed down our evaluation to two final candidates. We then implemented key aspects of our model in both platforms to compare them in terms of practical relevance. Our findings reveal that the Agents.jl platform outperforms the Mesa platform in terms of runtime performance, has a smaller memory footprint for large numbers of agents, and offers scalability, making it the most suitable choice for developing MAS for integrated energy and mobility models.

Keywords: Multi-agent simulation - Agent-based models - Mobility transition - Car-sharing