From CS 294-10 Visualization Sp10

Jump to: navigation, search



Group Members

  • Jon Barron


We plan to address the problem of automatic cartogram-style distance-based distortions of maps. Specifically, we hope to construct visualizations that will be helpful for bicycle navigation. We have previously developed a technique for routing cyclists based on *time of travel* and on *effort exerted*, which produces routes similar to those that a cyclist would choose himself. These routes are based on minimizing a cost function over routes, and this cost function is extremely confusing when visualized naively, such as with color-based encodings. For example, distance often correlates very poorly with cost (it is much more costly to go a short distance up a steep hill, than a long distance downhill). Our hope is to produce 2D distortion on these maps, such that distance encodes time+effort.

As our method for map distortion, we plan to use thin-plate splines for our distortion, and to hand-craft a loss function that effectively encodes cost as distance, while also producing a pleasing, and intuitive distorted map.

We expect to encounter difficulties with respect to asymmetry. The difficulty in biking from A to B is often very different from that of biking from B to A. How to best reconcile this issue is unclear.

Because our cost function is not monotonic in distance (it's easier to bike to the backside of a hill than to bike to the top of a hill, thus violating the triangle inequality) correctly encoding cost as distance should sometimes be impossible. We hope to be relatively robust to this issue.

Initial Problem Presentation


Final Deliverables



File:JonBarron-Code.zip - Run "go.m"

[add comment]
Personal tools