IWalkways is a tool to help facilitate good pedestrian walkway design. The primary emphasis of the system is to help the architect visualize and evaluate their proposal dynamically during the design phase. This prototype is written in Java for its ease of development and its portability. It has a simple design interface which does not constrain the designer, but rather it provides the architect with visual information to aid in their decision making. The system has design knowledge about the cost of building a walkway, how acceptably small the angle can be between walkways for a pedestrian to follow it, and the site. All of these considerations are displayed to the user when requested and most of them are customizable. This project was largely inspired by Kovacs and Galle's 1993 paper on walkway design, but it takes a drastically different approach in how it helps the designer solve their problem.
IWalkways differs from most other visualization systems for architects since it dynamically conveys visual cues during the design process rather than only working with completed designs such as (Wright and Hoinkes, 1993) and (McCullough, 1995). Not only will this allow the designer to see his or her potential problems at an early stage, it also might teach the user design heuristics. While designing, the user will be able see how different walkway layouts fair against the design criteria and hence should be able to learn through experimentation. This sharply contrasts with the papers by (Kovacs and Galle, 1993) and (Kovacs and Galle, 1994) since their logic program only determines if a human or mechanically created design is "acceptable". Although the underlying design heuristics of the software is similar between IWalkways and the proposed program that Kovas made (with notable exceptions noted in the aesthetics portion of the paper), how the software interacts with the designer makes the two programs radically different.
The logic program that Kovacs and Galle presented is exactly what it states to be-- a formal logic program. Although it is commendable that the authors explicitly encoded design knowledge which many other researchers have only theorized about, they probably did not package it in a way which is comprehensible by designers and more importantly are removing the designer from large parts of the design process. Kovacs and Galle made convincing arguments on the correctness of their loosely defined logic program, but it is unclear if these few logic rules can capture the full essence of designing. Unless these rules can capture all of the factors in design, they are at risk for producing inferior models. For example, if the quadrangle in question was overlooking a lake a designer might want to emphasis the number of walkways that run perpendicular to it. As Galle's system stands now, the designer will have to write first order logic rules in order to tell it to emphasize in some fashion the walkways leading to the lake or hope that the mechanical generative system will generate enough designs which might happen to include one which takes into consideration the lake. The former approach assumes that there exists an algorithm which can encode the designer's thoughts and further that he or she knows how (and wants) to encode them. The later approach may or may not lead to an acceptable solution-- if the problem was ill-defined to begin with, then only by chance will we arrive at a good solution. We can do better.
IWalkways adopts some of the design heuristics that Kovacs and Galles mentioned but presents them visually to the designer as guidance for design. These heuristics are just heuristics- never to be taken as solid truths. It is possible to generate an excellent design which does not meet all the design heuristics. IWalkways allows for heuristics to be broken, but at least the designer will be aware of potential consequences of their design (in this case, it might cost more to build, or that a pedestrian may be more likely to cut across the lawn). Also, IWalkways does not provide any generative support so it lays no claim that it is capable of designing on its own- no matter how many design heuristics it might know, it is still just a tool.