Tag Archives: Affordances

Dragonfly: New AI Agent engine based only on Perception

Here’s a preview of the completely new AI agent engine in Dragonfly. We’re really excited about it. For a couple reasons….

The first is that this AI engine is based only on perception, i.e. what the agent “sees” in a given frame. No navigation meshes, no looking around corners – just raycasting. This means that you don’t need to preprocess your geometry at all: simply drop in the agents and let them go….

The second is that they go up (and down…)  ramps! This may seem fairly trivial, but we actually found it to be quite tricky without nav meshes or preprocessing. How do you know what a ramp is? When do you go up one? When on one, why do you keep going?

Enjoy!

 

1 Comment

Filed under Architecture, Architecture in Combination, Dragonfly

New SPM Vector Components release coming up and a New Project…

Yes, look for a new SPM Vector components upload on Monday (maybe Tuesday…). We’ve implemented a Collision dynamic that works with any collection of surfaces, and a much needed scale and velocity fall-off dynamic customizing the granularity of the simulation. We’ve also broken out Acceleration as its own component, which cleans things up visually, and created subcategories in the Dynamics for 3d and 2d (any surface) forces…

AND, we’ve started up work on the Dragonfly project. I posted some screenshots of this component (isovist, intentionality, geometry simulation thing…) on the Grasshopper forum a couple of months ago, but I was never totally convinced that Grasshopper was the right platform for it. So, after much thought and discussion, we will be implementing it as a Rhino and possibly Revit plugin that calls up a standalone simulation environment. I’ll post some progress shots as soon as possible, but our release date is mid October.

Dragonfly will be presented at ACADIA 2011 in Banff, Canada.

Very excited!

Leave a comment

Filed under Architecture, Architecture in Combination, Dragonfly, Geometry

Buildable vs. Useable

A few years ago, a major driver for research in architectural geometry was the growing gap between what was imaginable and what was buildable. With big advancements in the design and construction of freeform surfaces, parametric models, and live physics engines, that gap has been considerably lessened. However, these days, I see the same kind of gap growing between that which is buildable and that which is useable.

 Our methods of evaluating the usability of designs, even if they are a result of real-time-sensory-input-fed-into-a-genetic-algorithm-that-optimizes-some-structural-property, are very much based on human intuition and previous experience. Now, I am very much in favour of preserving and developing the role of intuition and experience in this digital design age, but I can’t help but feel the similarities to the problems that engineers faced when confronted with a doubly curved surface not five years ago.

Motivated by this, Mike Braund, a Ph.d candidate at York University, and I have been exploring the potential of using ecological psychology to quantify, in some way, the usability of digital designs. So far, we have managed to implement a rudimentary version of ecological psychology into the Grasshopper environment as a proof-of-concept, but we need to seriously rethink the code in order to develop the idea further. Ultimately, we are trying create a simulation environment that can quantify and inform the design of architectural spaces based not only on the intention of the designer, but also the intentions of the user.

1 Comment

Filed under Architecture, Grasshopper

Density Plots

Here’s a plot of a 2×20 agents starting from different points, with different goals, interacting with a floor plan. When you take a way the plan, the density diagram is pretty interesting!

Leave a comment

Filed under Architecture, Dragonfly, Geometry, Grasshopper

Dragonfly: Architecture vs. Ecological Perception

Dragonfly is a first attempt at linking the theory of ecological perception to architecture. Given certain abstract programmatic events (e.g. view art, buy food, sit down, look up) and knowing roughly where they’re supposed to happen, Dragonfly simulates a user navigating the proposed layout. By observing the user’s behaviour, we can assess how effective the layout is.

Granted, in this example, the layout gets pretty chaotic, but it shows the principle…work in progress, folks.

Developed by:

  • Daniel Hambleton, Studio for Progressive Modelling, Halcrow Yolles
  • Michael Braund, Ph. D candidate, York University
  • Martin Walker, Studio for Progressive Modelling, Halcrow Yolles

3 Comments

Filed under Architecture, Architecture in Combination, Geometry, Grasshopper