**1. Introduction**

Agent-based simulations have proven valuable for studying traffic and mobility phenomena, including parking search and availability, without disrupting the actual traffic in a city. Many cities assign grea<sup>t</sup> importance to solutions to parking-related problems [1]. Those solutions include smart parking systems (SPS) which, mainly found as Parking Guidance and Information (PGI) systems, determine the parking occupancy and provide suggestions about parking availability [2]. SPS have been shown to have positive effects on driver parking success and on traffic flow [3]. SPS development has also significantly benefited from behavior models for parking search that help in analyzing the underlying phenomenon [4], and in testing of design and implementation choices [5,6]. Drivers are commonly represented as agents, using Agent-Based Modeling (ABM) [7,8], located in an urban environment. The environment can be represented using data from Geographic Information Systems (GIS), which provide a digital representation of the urban environment [9] and are already in place in many city governments.

When off-the-shelf simulation software is used to build a parking simulation linked to an SPS, the software often can impose data format and scripting restrictions that create conflicts with the SPS design and implementation. These conflicts prevent or complicate the SPS and simulator from sharing GIS data and algorithm implementation [5,9], affecting resource sharing and thus reutilization. Building or adapting a parking simulator so that it can share data and software components with its related SPS can benefit an SPS project beyond what is explored in previous parking simulations studies. Such benefits are important considering the increasing availability of city geospatial data [10] and the increase in governmen<sup>t</sup> interest in parking optimization [11]. For example, an SPS development

project for a city may benefit from (1) the incorporation of available geospatial services into the SPS development, and (2) the early evaluation of the SPS design and feasibility by means of a parking occupancy simulation.

This paper describes our experiences in building an agent-based parking occupancy simulator. The simulator had two major goals: (1) testing suitability of an SPS in the context of a university campus and (2) reuse of its development efforts (and code) during the SPS development. The SPS targeted by our simulator checks the occupancy of on-street parking spots using sensors, and handles logical spot reservation upon request. The simulator allows the exploration of parking occupancy patterns created by agents that either use or decline to use the SPS. Agents represent drivers of the most typical profiles of people who drive to and within the campus.

As a case study, experimentation was performed using the simulator to explore situations with different levels of parking demand and SPS usage. The results provide insights into a metric for SPS suitability evaluation from a driver's point of view. Also, the experiments allowed exploration of the reservation guarantee problem (someone stealing your assigned spot while you are en route to it), which arises due to the lack of a physical reservation enforcement. In summary, the main experiences and recommendations in this paper are:


To the best of our knowledge, no previous study has proposed the mentioned reutilization methodology relating an SPS and a parking occupancy simulator. The design proposal allows novel traits like running a simulation from current parking state data, or automatically using the latest environment information. Likewise, despite the fact that the reservation guarantee problem has been acknowledged by other studies, they did not study the problem incidence under several levels of SPS usage. Our analysis questions the acceptability of an SPS that promises a reservation to drivers and does not physically enforce the reservation. Furthermore, the parking studies mainly analyze parking search distance, which is only a part of the total driving distance. Additionally, the code of our simulator is freely available in a public repository.

The remaining sections of this article are as follows. Section 2 presents relevant previous studies and supports our design considerations. Section 3 describes the relationship between the SPS and the parking simulator. Section 4 explains the simulator's details. Section 5 describes the case study of the simulator for the experimental evaluation of SPS usage. Finally, conclusions and acknowledgement sections are presented.
