Week 6 – Starting the Simulation

Edward Z -

Greetings, and welcome back!  This week, I’m going to be going in-depth about how my simulation works and how I plan on making it.

A few days ago, I was able to meet with my faculty mentor and discuss how the simulation pipeline works,  including the overall framework and the simulations’ key parameters. I’ll try not to make this post too long, but I want to include some more technical information to keep you all informed.

The simulation will be broken up into 4 sections:

1) Problem statement: Our goal is to minimize queueing times in any school using a smart reservation system – a separate lane/entrance dedicated to only scheduled arrivals (as discussed last week). We are given current arrival and departure times and our priority is to calibrate the waiting time, minimize it, and then find a new time horizon.

The waiting time has two components. On the one hand, we want to minimize the parent’s idling time, and on the other, we want to minimize the student’s waiting time outside. We’ll assume that all agents will maintain user equilibrium – or in other words, the very best selfish choice for their own benefit. After defining the constraints, we can set a utility function based on these parameters: Uand Us, for students and parents respectively. The utility function presents a tradeoff between the preferred arrival time (assuming the linear function increases due to less queueing volume) and the queueing delay function (assuming the linear function decreases due to fewer cars arriving later and later). Thus, our function for each utility becomes U = D(delay for vehicle at time t) + β (utility coefficient to calibrate differences in value of time between vehicle and student waiting times)* PA(preferred arrival at time t) + W(waiting time of the student at time t). This essentially describes the agents’ aggregated behavior when choosing the viability or necessity of using the reservation system.

By using this utility function, we can then attempt to optimize the system by introducing the 3 pickup systems described before. There’s quite a bit more to it, but I think I’ll leave it at this for now. In the final research paper, I’ll be going into more depth regarding pickup time deviations, preferences, user equilibrium settings, etc.

2) Dynamic Traffic Assignment: After defining the problem, we optimize it, and then generate the optimal trip demand based on these constraints. This will produce an Origin-Destination Demand Matrix (OD demand matrix) which will describe the paths vehicles will take based on the behavior and design of the system. It’ll generate a mesoscopic (i.e only lanes no lane changing) network and a file will be produced: the agent.csv. Using agent.csv we’ll then do the next step.

3) Connected and Automated Mobility (CAM) System LITE: In this step, the goal is to create highly granular lane-changing arcs within the pickup area. Meaning, the cars will be able to change lanes and conflict points will be created to add realism to the simulation.

4) Simulation Visualization and possibly 3D: after all situations are simulated, it is crucial to be able to convince decision-makers, planners, and school administrators that it might be worth it to reduce a lane’s capacity for the overall system benefit. Thus, the final plan is to make a 3D visualization, with my original experience in Unity. However, if the time comes up, and the project cannot go through with this, I’ll hold onto this idea for a “future” goal as I continue to improve it over time.

By going through this pipeline, we can ensure that the simulation is accurate not only to the problem but also to its beneficiaries. I expect this to take the remainder of the month, where I could quickly draft a research paper and poster to make the presentation on time. While I do feel I’m a bit late and stressed, now that I have a clear step-by-step plan, it’s gotten much clearer.

Finally, to end off this enormous blog post, a few things have changed:
1) While simulation is multi-scale, it will only focus on one school at a time.

2) The strategies used will remain as simple as possible, meaning space or capacity will be allocated. No rewards or punishments will be used.

3) A survey will still be asked (at least for now, might change). The project has gotten much more straightforward and less open-ended.

Again, I’ll be going over this in more depth in the research paper. But for now, I’ll see you all next week. Thanks for reading!