Archive for June, 2006


AlifeX was completed a few weeks ago, and its been quite a haul since. Some thoughts from then and now…

Update: Some of the Keynote Talks from AlifeX are now online.

It was incredibly stimulating to jump in the deep end of this research environment, so even though I’ve been exhausted I’m finding it hard to sleep. Pages of my sporadic notes link to lifetimes of research directions. I was especially taken with Evolutionary Development work, and Niche Construction, and naturally spatial relations, and think evo-devo may be key (along with community interaction I explored, and hope to explore further, in my work) to the question of Evolution of Complexity explored on Saturday. Evolution isn’t a blind search, its embedded in history and physics and chemistry, some pathways are quicker to climb in the evolutionary landscape, and sometimes some process is hit apon that spontaneously leads to leaps. And sometimes what makes most sense just won’t happen .. cause while it would be really cool to be a Centaur, our four appendage line makes this pretty unlikely. It’s a Hunch not Hypothesis .. need to find more of those elusive Hypotheses.

Space really needs some highly measured tests of its affect on simulation. So many ALife models incorporate space, with agent based modelling being perhaps a core definition of ALife traditionally, yet what is the precise effect? My idea is to implement 2-D Lotka-Volterra, varying the spatial heterogenity from perfectly homogenous, with results in line with equation based models, gradually to different spatial distributions and measure some systematic properties.

Another big question to me is biological relevance. I get a spectrum of opinions on how real ALife should be. There’s the mantra of simplicity, and the frustrating alure of incorporating real data.

I spent a lot of the conference looking for models of organism development supporting interesting, simple, possibly hierarchical and semi-open ended development (towards the ALife holy grail of open ended complexity growth). What happens when a model like this evolves in a food web. James Crutchfield’s epsilon machines are designed for a very different purpose, but something about coupled finite automata seems suitably abstract and potentialled for this. Finite automata have produced interesting results for me before, in evolving swarm foraging .. is there a way to make FA hierarchical? Wriggraph, while not suited for my purposes, was a very interesting model that combined ontogeny and motility in a single underlying mechanism based on chemical diffusion. Often for me, misunderstanding an idea leads to a new research question: can an L-System-like-System self modify its rule set to communicate some notion of position along the growth path; can this lead to arbitrary L-System growth, with Growth Universality?

The most ALifey, mind blowing work I saw at the conference was from the Cornell Computational Synthesis Lab. I’m not really into robots, but this stuff makes me wish I was. These are the same folks that built the Golem “Self-Assembling” Robot a few years ago. Their new work is just as great. A starfish like robot (with only four appendages, which starfish have five .. how is radial symmetry related to the fibonnacci sequence again) is placed in the world with no conceptions of its own self, only some methods to make use of experience. It builds up a model of its own body by iteratively generating models of itself, and testing those against its experience using this model .. an “Emergent Self-Mode” perhaps related to development in animals, children .. and eventually successfully evolving a method of motility. Chris Adami asked “Is this dreaming?”, similar to how amputees will have dreams of missing limbs, they are updating their internel model of self. Self-assembly is being pushed further, with voxels suspended in oil building up structures based on L-System like growth; each voxel can be quite sophisticated, for instance even a 300 micron tile could host a web server.
Incredibly, CCSL is pushing the frontiers of material science as a side project. The Golem Robots were a slight overexageration; wiring and joints were manually introduced post-printing. Now, they have developed their own Rapid Prototyping 3-D printer which can incorporate multiple materials, and have been working on printable materials that can store and conduct energy. That is, a 3-D printer that can produce a complete robot including wiring and batteries!! And the price of these printers they’re developing are potentially dropping below $1000. This is close to Fab@Home surreallity. I feel like I saw the future before anyone had even thought about it.



The team at Lokku have done a terrific job with the launch of nestoria. I’ve been advising them a bit on how to implement geographic search and integration of various geodata sources, and it’s great to be associated with them. If web cluefulness and usefulness is any guideline, they’ll soon be a force on the London (and later on UK, and Europe) real estate market.

Too often, the innovation on the web, and especially the geospatial web, is reflected commercially only in the Estados Unidos. There are reasons for sure, like the traditional lack of free geodata, but with OpenStreetMap and Google Maps relentless expansion, the playing field is levelling quickly. Trulia are a natural comparison, in the Web 2.0 of real estate, all basking in the glow of housingmaps but here I think simplicity is key. What will be a challenge for nestoria is maintaining the current simplicity as the feature set grows. Especially cool now is the neighborhood info — transport, schools, doctors, and of course pubs.

It’s good to see commercial ventures make use of free geodata. Hopefully, nestoria will soon be reciprocating with GeoRSS enabled feeds. They are already wise to benefit of engaging with the open source community. Mapstraction, the abstract interface to all the big mapping APIs we’ve been developing, has received its initial support from nestoria.


Santo Domingo

My current location


OpenStreetMap knows the way to San Jose

The mission was to map out a piece of the geodata netherworld .. a US location that Navteq/TeleAtlas had not yet driven, and distributed out to GYM. Preferably in San Jose, host to Where 2.0 2006 🙂

San Jose, largest city in Silicon Valley, is a rapidly sprawling suburban metropolis. This unsustainable mode of development is generally considered detrimental to open space, consumed by low density housing, energy resources, squandered on car dominated transportation, and community, split apart by satellite dishes and 6 lane roads. But what is bad for the environment is still brilliant for OpenStreetMap! With all the new development, there was bound to be some corner of San Jose unmapped. And ok, I actually found it to be a very interesting place.
A quick search brought up the Development Activity Highlights and Five-Year Forecast, published annually by the San Jose City Planning Department. This comprehensive document featured several helpful maps, with areas planned for development clearly highlighted. The Evergreen area, on the eastern foothill slopes, had several promising, large, yellow areas.

Also at this time, Ben Gimpert was importing the entire Tiger dataset into OpenStreetMap. Tiger, published periodically by the US Census, contains streets for the entire US. With the addition of GPS derived streets, OpenStreetMap could potentially have the most up to date map of the area. Ben helpfully loaded in Santa Clara County ahead of schedule (it was working on Alabama at the time, then Arkansas…).

Just past Evergreen College, the plan was all on the ground.

Development Proposal

The first zones had only the barest imprint of a proto-suburbia. Possibly last week this was open pasture land, now with a gravel base road and basic road drainage. This development probably won’t complete for a year or two, and sometime after that it will appear in Navteq. Yet, this place is very important today, if you are concerned about suburban sprawl or saving up for a new home.

Your future home?

Just down the road was the first big error in Navteq. Pre-development, Fowler Rd connected to Yerba Buena Rd. The last segment of this connection is now blocked off, with traffic flow directed down the newly constructed Altia Ave. Not everyone is so happy about this, evidenced by the smashed fence on one side of the segment.

No thru road

Before looking at the results further, let’s examine what’s currently available from G-Y-M. All three source street data here from Navteq. Microsoft and Google both have aerial imagery sourced from the USGS and are basically identical. In the center of the imagery is a large plot of undeveloped orchard, and in the upper left corner more land under development. Yahoo!‘s imagery is more recent, with lines of homes eating away at the remnants of the orchard, and playing fields completed at the new high school. The pace of development is so fast in this area, it may be possible to date the week these images were taken. When I visited, the orchard was almost entirely gone, replaced by Orchard Heights, with families already moved into most of the houses.

GMaps Evergreen

Yahoo Evergreen

And some digression into suburban lament. Silicon Valley was once the Valley of Heart’s Delight. This was some of the most productive agricultural land in the world, full of cherries, walnuts, apricots … now the land itself imprinted with circuits, sprawling in some kind of reverse Moore’s Law. The remnants of the orchard, machinery and boxes, were stacked, rusting and weathering just over the fence from the new neighbors.
Old Farm Equipment and New Houses

The new residents of Orchard Heights seemed to be Silicon Valley economic migrants, Phillipino, Chinese, and Sikh families drawn to high paying chip design and programming jobs. They seemed confused, not only by the strange presence of a GPS toting dude taking photos of street signs, but by their own presence. People stood and conversed on median strips and corners of small lawns, or paused and pondered utility boxes and fresh landscaping. The space had very little preconceptions, patterns of behavior, or delineations among the subtley similar New Urbanist flavored homes. This place was all dream, waiting for some history, oil drip, or odd garden sculpture to bend the neighborhood into some kind of story.

Thing is, most every SUV parked in the driveways had GPS navigation. All the tools are nearly present for these streets to exist in our popular tools. Certainly the developers have interest in enabling driving directions for new home buyers. And is it even possible that the act of mapping might be the first step in connecting people to their new home, moving from space to place? I certainly had my eyes open.

Here’s a typical OpenStreetMap photograph. The corner of Botticelli and Michelangelo. Just maybe the developers had Italian ancestory?
Botticelli Dr and Michelangelo Dr

I took photos at every intersection, and geotagged their EXIF headers with WWMX Location Stamper, which simply matches up the timestamp in the GPS track log with the photo timestamp. WWMX overlays the whole thing on Microsoft’s Navteq sourced streets, just for presentation in that software. I’m showing it here just for demonstration; OpenStreetMap bases its data completely on open and free data sources. My tracks fit like a missing suburban puzzle piece.
WMMX Evergreen

The traces fared less well when layered over the existing Tiger data in OpenStreetMap. A number of existing streets in the area don’t align with the tracks, and even less data than Navteq is available overall. Tiger must have sourced this data from previous development plans, prior to actual building which have since been modified. To map this area accurately, we’ll probably need to go in and do another round of data collection. No biggee, just another afternoon will probably do it.

The take away for me is twofold. Not only will collaborative OSM data collection soon be competitive with commercial data collection, it will be a means of fostering community in an ever increasingly mobile world.

Comments (7)


I’m in Bloomington Indiana for AlifeX. The Evolution of Complexity Workshop is tomorrow, I’m presenting my work on evolving food webs.


Java Quake

WorldWideHelp and ShelterFinder are stepping up in response to the Indonesian Earthquake. I’m engaging, trying to free up the flow of satellite imagery, village/hospital/airport location dbs, mapping apis, photos, textual reports, etc into a fluid coherent shape, meaning GeoRSS and KML.

First pass, I took locations listed in the ShelterFinder wiki, and manually created a quick GeoRSS file, and loaded into a Yahoo Map. Of the big 3, Yahoo currently has the best map coverage of Indonesia. From this, I experimented in mapufacture, layering weather and disaster alerts with these locations in the mapufacture JavaQuake.

Later I heard about satellite imagery released by the German Center for Satellite Based Crisis Information, taken post quake with damage assessment overlays. This would be very useful in Google Earth. Helpfully, these were posted with georeferencing world files. These images were projected in some form of UTM, while Google Earth expects image overlays unprojected with lat/long extents. The projection itself was not specified in the world file, but directly in text with the image as “UTM Zone 49 S”. I hopped on irc #geo, and was quickly aided by Tyler Mitchell, Schuyler Erle, and Chris Schmidt. After some experimenting, the method hit on was to use gdalwarp to convert each jpg/worldfile into an unprojected GeoTiff, and then gdal_translate to produce a jpeg compatible with Google Earth, along with another worldfile listing in lat/long. Here’s the geek…

gdalwarp -s_srs '+proj=utm +zone=49 +datum=WGS84 +south' -t_srs 'EPSG:4326' bantul_high_1.jpg bantul_high_1.tif
gdal_translate -outsize 50% 50% -of JPEG -co WORLDFILE=YES bantul_high_1.tif bantul_high_1_50.jpg

Thanks to Chris Schmidt for doing the heavy lifting here. I took the results, calculated the extents, and produced a KMZ of remote assessment of Indonesian Quake Damage.
Overlays of Indonesian Quake Damage Assessment

Still a year or so into map mashup hype, I’m surprised this wasn’t easier, and that the original publishers didn’t think of producing KML for their imagery. And there’s loads of other sources producing images and data that could be useful for relief efforts, and put into KML etc. There’s still a ways to go to design geodata for sharability.

A side note, at XTech the other week, Anselm Hook and I were discussing methods to specifiy image overlays in GeoRSS, and I think we found something that could be useful. An image can be specified using MediaRSS, and its extents specified with a GeoRSS Box Geometry, with a relationshiptag of something like “image-extent”. Mappers and visualizers would treat this the same as a KML GroundOverlay. Why not just stick with KML? Cause GeoRSS already has wider support, and greater potential for support and aggregation, and there are tools for translating GeoRSS into KML, if not the possibility for native support at some point.