Berkeley Delivery Simulation

Simulation-based digital twin for complex real-world traffic modeling

Using HASH, students from the University of California, Berkeley, have built a traffic simulation of delivery-cars in the city of Berkeley to predict the efficiency of delivery services and the impact on street congestion.

Delivery cars (in blue) go from the restaurant and stores (in red), to the households (in yellow), according to a Poisson distribution of needs._

The model simulates deliveries in the Berkeley area through a combination of real world data and customizable agents:

• →

  Store locations are scrapped from Yelp

• →

  Home locations are generated from a random distribution with family sizes pulled from the census data.

Delivery agents move between locations, responding to requests from homes and stores. This results in increased traffic, the costs of which are captured and displayed. The scalability and flexibility of the model allows for simulations to be generated for different cities easily.

Source: Anton Bosneaga on YouTube (2020).

Simulation driven modeling of traffic congestion and navigation routing is a promising approach to more robust traffic prediction, and a potentially high leverage tool for city planners and companies to model what-if scenarios.

Credit: Anton Bosneaga, Jackson Le, Malo Le Magueresse, and Peter Zhu.

---

Update 2024-02-05: This post was updated to remove a broken link to the old hashaiproject.pythonanywhere.com website, which is no longer live, but which had previously contained a writeup and more details about the project.