Group Project: Generative Topology for Participatory Redesign of Neighborhoods

The purpose of this project is to consider a particular novel direction for generative AI beyond that is linked to an application we explored in the main part of the course.  It is to be done in a group of roughly 3 people: please coordinate group composition yourselves.  All students in a group will receive the same grade.  Although implementation of data processing and algorithms is certainly encouraged, it is not strictly required.  If you look at the grading rubric below, it is focused on problem formulations and discussion of possible techniques. 

• The deliverable is a 2-4 page paper in IEEE two-column format.  

The paper is due May 2 at 5pm via Gradescope.

The problem to be addressed in the project is of generative neighborhood topology to support participatory redesign of neighborhoods and is inspired by the guest lecture of Luis Bettencourt and related resources:

• C. Brelsford, T. Martin, and L. M. A. Bettencourt, "Optimal reblocking as a practical tool for neighborhood development," Environment and Planning B, vol. 46, pp. 303-321, Feb. 2019.
• C. Brelsford, T. Martin, J. Hand, L. M. A. Bettencourt, "Toward cities without slums: Topology and the spatial evolution of neighborhoods," Sci. Adv., vol. 4, eaar4644, Aug. 2018.
• F. Schembri, "Urban planners aim to eliminate slums—with computer programs," Aug. 2018. 
• Million Neighborhoods Initiative, Mansueto Institute for Urban Innovation, University of Chicago.

The basic question posed to you is to formulate a generative AI problem for the redesign of neighborhoods so that the community is given a diverse set of generated possibilities that may improve quality of life, and suggest algorithmic methods to do such generative topology design.  In order to facilitate your thinking and also to allow you to try out some ideas with algorithmic implementations, some sample data is given at this Box link, which has a zip file.

The primary file is kblock_full_freetown.geojson, which is a street block level file for Freetown Sierra Leone (see bullets for column descriptions and the image below). There are also 5 separate files with building geometries that fall within the 5 districts covered in the former file (these files are named "buildings_SLE_*.geojson"). The last file is called osm_streets.geojson and contains linestrings corresponding to linestring features including roads, coastline, waterways, etc.

The file called "kblock_full_freetown.geojson" contains street block level geometries with the following columns:

• block_id: unique ID for street block unit
• gadm_code: district level geographic ID. There are four districts in this dataset: "SLE.4.1.4_1" "SLE.4.1.2_1" "SLE.4.1.1_1" "SLE.4.1.3_1" "SLE.4.2.1_1"
• country_code: country code for Sierra Leone
• block_area: area of block unit in meters square
• building_area: area of all buildings in block unit in meters square
• building_count: number of buildings in block unit
• building_layers: number of buildings in each k-complexity layer
• k_complexity: integer value corresponding to level of informality
• geometry: geometry in WGS 84 CRS 4326

The grading rubric is as follows.

Background (4pts)

• (4 pts) Well-organized, coherent, and well-written description of urban planning background (including value of participatory design and need for diverse generation therein). Includes relevant background on topological characterization/optimization of neighborhoods. Gives appropriate motivation for performance metrics to be used in algorithmic approach.
• (3 pts) Reasonably accurate (some minor errors) and reasonably complete.
• (2 pts) Inaccurate, missing main ideas/points.
• (1 pts) Extremely limited.
• (0 pts) Not included.

Algorithmic Approach (4 pts)

• (4 pts) Contains a complete, precisely specified, and insightful approach to the topology generation problem for neighborhood redesign.  Provides motivations for all design choices, including performance metrics.  Gives some indication that the algorithmic approach will work well (theorems, simulations, analysis of real data, etc.)  
• (3 pts) Largely complete and correct, but missing some main element listed above.
• (2 pts) Partially complete and correct, missing several main elements listed above.
• (1 pts) Incoherent, incorrect, or otherwise very limited.
• (0 pts) Not included.

Paper Interestingness and Overall Impression (2 pts)

• (2 pts) Interesting and fun to read.
• (1 pts) Either uninteresting or not fun to read.
• (0 pts) Not interesting and not fun to read.