Optimized Bridge Island
Optimized for leisure, for convenience, for safety
Eric Chyou, Hao Zhong, Junho Lee, Kyounghwa Lee, Seokhyun Kim, Sunghyun Kim | Generative Design | Fall 2021 | GSAPP
| Current Issue_ Hudson River:
Currently, there is no infrastructure connecting directly from New Jersey to Manhattan from 59th to 125th street. Our goals are to minimize travel distance between the two, help create more enjoyable spaces, and give the city more room to breathe.
We are designing an Island Bridge through simulation and optimization utilizing a combination of Grasshopper and Python to generate different iterations.
We are to identify connection points through traffic simulation, minimize travel distance between programs throughout the Island Bridge through optimization, maximize a leisurely experience for the commuters; all while minimizing the impact on the surrounding context such as height limitations, maritime travel restrictions, and others. The Island Bridge would offer areas for various programs such as clinics, after-school programs, dining, and leisurely programs that would benefit the close-by communities.
| Methodology:
1. _Design Space Model
The 45 by 7 grid is defined on the Hudson River and three points are defined on each side of Manhattan and New Jersey. One point is chosen from each row of the grid. The connection between these points is generated by Delaunay triangulation. The points and connections are soon transformed into islands and bridges using parametric design.
_a. Land Points and Grid
We started off by identifying several landing points on both riversides of the Hudson River. Based on the areas that have been determined, a coordinate system over the water for the Islands’ initial points is generated.
_b. Island Points
With the Grid generated, Red Points that locate the potential Island locations are randomly generated by Grasshopper. Out of the 45 X 7 possible locations, one point is picked out of each row. However, in certain instances, less than 7 points are chosen in order to optimize the travel time.
_c. Analysis
The Red Points are then connected to each other to generate the pathways/Bridges that eventually connect the two states. A starting point is selected and can be assigned to various locations for optimization. The possible routes between programs and closest Island locations are suggested by Python. These routes are then calculated for their travel distance.
Meanwhile, the Bridges are computationally generated according to the pathways generated through the previous series of operations.
_d. Output
With the travel time determined for each of the possible routes between various programs and Island locations, the data is then inputted into Discover for optimization. The output from Discover would feed the data back to the Grasshopper definition and find designs with the most efficient travel time. Through a series of iterations, the overall definition would propose the optimized design for the project in order address the issue.
2. _Input & Output Parameters
The area of the site was identified and with that the coordinate system has been laid out to cover the Hudson River. The inputs of the grid could be parametrically generated.
The Blue Points/ Landing Points were manually (manual inputs) identified and were assigned at various locations on the Grid intersections across both riversides.
Since there are 7 rows and one point out of each row was proposed by Grasshopper; consequently, there are seven potential inputs (Island points). The inputs are seven integers between 0 to 44 and the total possible number of inputs is 4⁵⁷. As mentioned, these integers are interpreted as points and soon transformed into Islands.
These Island points are inputted into line-generating components to create pathways/Bridges. Meanwhile, the possible routes traversing from various programs to Islands are generated and calculated through Python. The lengths of the pathways in combination with the lengths of the proposed routes are translated into traveling time, these output numbers are then fed and analyzed by the travel-time Python script.
These analyzed input numbers into Discover for comparison between other potential sets of numbers/lengths of travel. The outcome of this optimization operation would yield new sets of points across the 45X7 Grid prompting new Island Points/locations. These new sets of points are then fed back to Python and analyzed again for their new traveling time/numbers.
Through generations of iterations, the definition is programmed to determine the outputs of Island points/locations optimized for the shortest/most efficient routes (lines) connecting from programs in New Jersey to Islands to between-Islands to programs in New York. Consequently, these series of poly-lines are inputted into a series of operations to generate the final geometry.
The generated poly-lines and offset poly-lines are fed into a series of geometry-generating operations: extrude, loft, pipe, etc. Finally, through the numerous operations, as an output, the definition yields the optimized Geometry.
3. _Performance Metrics
Since we are tackling the accessibility to infrastructures and we take into account the number of choices one gets is important, our main metric is average travel time to 20 locations nearby. Due to the size of the model, we used a Python package named ‘networkx’ to analyze the shortest paths to calculate the travel time to the various nearby programs/infrastructures.
| Conclusion
Some of the limitations we have encountered are the boundaries of the design being between George Washington Bridge and Lincoln tunnel. The project’s limited scope of analyzing the travel time to the infrastructure, so many were not accounted for such as the impact on its neighborhood or increased demand caused by new bridges.
Some of the intuitive solutions in our design space would be to evenly distribute islands between George Washington Bridge and Lincoln tunnel. As expected, the optimal solution that came out from Discover was the one with relatively more evenly distributed islands.
Since the generated design was a network rather than a single connection, the differences between them are subtle. However, since the metric (average travel time to nearby infrastructure) is clearly measurable, the optimal solution was searchable. Moreover, there are other metrics that can be used later such as the number of infrastructures that can be reached in 20 minutes. If one combines these metrics with the original metric, more effective solutions can be found.
