Connect your existing design models
Discover is designed to connect seamlessly to your Grasshopper models, with minimal overhead required for taking an existing model and setting it up for optimization. A small set of components handle connecting your model to the Discover server, and specifying the inputs and outputs passed between your model and Discover. Meanwhile, all optimization and visualization tasks are handled in an external, web-based platform.
Generate input parameters
Discover supports a range of input parameter types to drive your generative models. This includes Continuous parameters for variables that vary continuously, Categorical parameters for values that take on one of several defined options, and a special Sequence parameter which defines a unique ordering of a set of items.
Define objectives and constraints
Discover supports both single and multi-objective optimization workflows, and has full support for constraint handling. Objectives and constraints can be easily defined by connecting output values from your models into a special set of Objective and Constraint components. Context menus within the components allow you to select the type of objective (minimize or maximize), or the type of constraint (less than, greater than, or exactly matching a specified value) you want to use.
Once you’ve specified the inputs, objectives, and constraints you wish you use in your model, you can use Discover’s web-based interface to run optimization jobs. The Run Options panel allows you to specify the parameters of your optimization job, including the number of designs to create in each generation, the number of generations to produce, and the mutation rate to use when generating new designs.
Reveal complex trade-offs
Discover’s visualization interface is based around a 4-d scatter plot which you can use to monitor an ongoing optimization or to evaluate the results of a previously run optimization. Each valid design is displayed as a circle, while designs that violate one or more constraints are displayed as hollow squares. The Explore panel allows you to map any objective value to the four dimensions of the scatter plot (x-coordinate, y-coordinate, circle size and color). You can also visualize the design or generation number to evaluate optimization trends over time.
Explore a wide space of design
Discover supports a bi-directional data flow with your Rhino Grasshopper model. Once an optimization is run, you can use the scatter plot to explore the set of designs in your design space, filter to show only the best-performing designs, and select a set of designs for further analysis. Clicking on any design in the Designs panel will automatically regenerate it in your Grasshopper model so you can proceed with the rest of your design workflow.