> On 9 Dec 2016, at 18:20, Bin Jiang <[log in to unmask]> wrote:
>
> Alan,
>
> If the EVAS agents is geometry aware, they cannot be long sighted. Let
> me explain why, and correct me if I am wrong.
>
> If we put a space into a 100 by 100 grid, then the visibility graph
> would be derived from the 100 by 100 grid. In other words, this
> visibility graph contains 10,000 nodes, and their mutual visibility as
> links.
That is correct. Then on top of this at each of the 10000 nodes the links to the others are sorted into 32 radial bins. In this way the lines of sight from each node are contained in the representation in all directions. An agent on some node with some heading can therefore look up all the nodes that are within its direct visual field. The stochastic search involves the ages selecting a node visible from where it is at random from those it can see, turning towards it and moving forward three steps (you can select how many, and of course this is scale dependent, but lets say 3) . It is now on a different node with a different heading. It repeats the process - select at random a node from those it can now see, turn and move. The trails of where many agents have moved according to this algorithm is recorded.
They are long sighted in that if the space is uninterrupted across a whole system they can ‘see’ across it and may select a node on the far side as somewhere to head towards. They are geometric in that they have a representation of the full geometry of the system at their disposal (represented by the VGA).
Alan
>
> In the PageRank inspired agents, we used axial lines induced
> connectivity graph. This connectivity graph is much more economic than
> your visibility graph; your visibility graph is geometry or grid
> oriented, while my connectivity graph is axial lines induced.
>
> Thanks and cheers.
>
> Bin
>
> On 12/9/2016 5:49 PM, Penn, Alan wrote:
>> Bin,
>>
>> the EVAS agents are long sighted and geometry aware with front facing vision. In effect they are a stochastic search of the visibility graph by an anisotropic agent with forward motion and vision. Both motion and angle of vision is adjustable (motion in terms of number of ‘steps’ between random reelection of the next heading/direction of travel; angle of vision adjustable between tunnel vision -one bin- and 360 vision -32 bins). For each grid point in the VGA all visible VGA grids are precomputed and sorted into 32 heading bins. Thus the space available that any agent can see, given its heading and location is simply stored in a look up table. This allows a stochastic search of this structure to be carried out for multiple independent agents in very fast time. The key thing about this is that the results although stochastic are completely deterministic. It is a Markow process without feedback or memory (in its simplest form). In other words it is another form of analysis of space like all other space syntax measures with the interesting property that the representation captures in an elegant way the anisotropic nature of foreword facing vision.
>>
>> Thus Farida’s observation of a difference between VGA integration and EVAS agent flows is mainly a product of the long aisle lines and the shorter cross axes. The later are more integrating because each connects to all the long aisles, but the long aisles attract more agent movement.
>>
>> Alan
>>
>>
>>> On 9 Dec 2016, at 14:02, Bin Jiang <[log in to unmask]> wrote:
>>>
>>> Hi, Alan,
>>>
>>> Do you think these agents are short sighted or geometry oriented? They
>>> determine their move in a step by step fashion. In our agent-based
>>> simulations, which were inspired by PageRank's random surfer model, the
>>> agents are long sighted or topology oriented.
>>>
>>> https://www.researchgate.net/publication/45878700_Agent-Based_Simulation_of_Human_Movement_Shaped_by_the_Underlying_Street_Structure
>>>
>>> Just an idea out of the discussion. Your comments in particular.
>>>
>>> Cheers.
>>>
>>> Bin
>>>
>>> On 12/9/2016 1:40 PM, Penn, Alan wrote:
>>>> Farida,
>>>>
>>>> this is exactly what we found in the analysis of the IKEA showroom floor plan, VGA quite different to Agents. The difference is explained by agents having forward facing vision and a process that leads to them moving towards open space. What IKEA designers had done was to ‘hide’ shortcuts by always hiding them behind partitions so that with forward facing vision you would have turn round to see the shortcut. Agents tend not to turn round but keep going forwards and do miss the shortcuts. This difference leads to a completely different analysis result. in VGA you can set the angle of vision number of bins to 32 - i.e. to 360 degrees and then the result approximated vga integration more closely.
>>>>
>>>> Alan
>>>>> On 9 Dec 2016, at 03:49, Farida Nilufar <[log in to unmask]> wrote:
>>>>>
>>>>> Dear All,
>>>>>
>>>>> I am using Agent-based Simulation to find av evacuation scenario for a factory. So far I understand Axial & VGA analysis gives an understanding of configuration and in case of simulation, agents follow the clues of the morphology/configuration. In one case Agent simulation (any point) appears totally different from Axial/VGA (Integration) results. Agents prefer a route which can not be explained by any means. Why?
>>>>> I am attaching
>>>>>
>>>>> Prof. Farida Nilufar PhD
>>>>> Department of Architecture
>>>>> Bangladesh University of Engineering and Technology [BUET]
>>>>> Contact: 0088-02-9665650-80/ 7221 or 7153
>>>>> email: [log in to unmask]@arch.buet.ac.bd
>>>>>
>>>>> <Needle Drop Ltd._Any Point_Agent Simulation_30.11.16.graph><Needle Drop Ltd_4th Floor_DXF_30.11.16.dxf>
>>> --
>>> --------------------------------------------------------
>>> Bin Jiang
>>> Division of GIScience
>>> Faculty of Engineering and Sustainable Development
>>> University of Gävle, SE-801 76 Gävle, Sweden
>>> Phone: +46-26-64 8901 Fax: +46-26-64 8758
>>> Email: [log in to unmask] Web: http://fromto.hig.se/~bjg/
>>> --------------------------------------------------------
>>> Academic Editor: PLOS ONE
>>> Associate Editor: Cartographica
>>>
>>> BinsArXiv: http://arxiv.org/a/jiang_b_1
>>> Axwoman: http://fromto.hig.se/~bjg/axwoman/
>>> ICA: https://sites.google.com/site/commissionofica/
>>> Geomatics: http://fromto.hig.se/~bjg/geomaticsprogram/
>>> RG: https://www.researchgate.net/profile/Bin_Jiang3
>>>
>>> [Högskolan i Gävle]
>>>
>>> Högskolan i Gävle, 801 76 Gävle • 026 64 85 00 • www.hig.se<http://www.hig.se>
>>>
>>> För en hållbar livsmiljö för människan
>>>
>>> University of Gävle, SE-801 76 Gävle, Sweden • +46 (0) 26 64 85 00 • www.hig.se<http://www.hig.se>
>
> --
> --------------------------------------------------------
> Bin Jiang
> Division of GIScience
> Faculty of Engineering and Sustainable Development
> University of Gävle, SE-801 76 Gävle, Sweden
> Phone: +46-26-64 8901 Fax: +46-26-64 8758
> Email: [log in to unmask] Web: http://fromto.hig.se/~bjg/
> --------------------------------------------------------
> Academic Editor: PLOS ONE
> Associate Editor: Cartographica
>
> BinsArXiv: http://arxiv.org/a/jiang_b_1
> Axwoman: http://fromto.hig.se/~bjg/axwoman/
> ICA: https://sites.google.com/site/commissionofica/
> Geomatics: http://fromto.hig.se/~bjg/geomaticsprogram/
> RG: https://www.researchgate.net/profile/Bin_Jiang3
>
>
> [Högskolan i Gävle]
>
> Högskolan i Gävle, 801 76 Gävle • 026 64 85 00 • www.hig.se<http://www.hig.se>
>
> För en hållbar livsmiljö för människan
>
> University of Gävle, SE-801 76 Gävle, Sweden • +46 (0) 26 64 85 00 • www.hig.se<http://www.hig.se>
|