Part 1 - Optimal Routes for Canoe Travel

In the first part, I calculate and evaluate optimal routes for canoe travel. Together with Jan Hildenbrand, a master student, we developed a software that calculates pathways between the islands. We are taking into account environmental data from weather models describing currents and wind conditions. This model has been used for a number of case studies in the Lesser Antilles by Emma Slayton. In order to validate to what extend humans were capable of learning to find optimal pathways, we developed a paddling  game for mobile devices together with Roman Lamsal, a bachelor’s student. This game runs on mobile devices and can simulate voyages between the islands based on real-world current data. We are currently setting up a user study to test how well players can adapt to the currents and find the optimal routes without actually seeing the current data.

Part 2 - Ceramics

In the second part, I examine assemblages of ceramic sherds from various excavations in the Caribbean. For each sherd that is found at a site, various attributes are recorded in a database. Based on these attributes, I generated perception tests for ceramics experts that allow me to calculate a matrix of perceived distances between the possible values of each attribute. Based on these distances, the sherds can be clustered into a set of types. The amount of sherds of each type can then be compared between the sites using various measures, resulting in a similarity network. The network can then be used to validate (or invalidate) the distances generated through the canoe simulations.

In joint work with Lewis Borck from the University of Arizona we developed an additional perception study based on pottery in which we ask archaeologists and indigenous potters from the US Southwest to arrange real archaeological sherds according to some given criteria. Comparing these arrangements, we hope to be able to quantify potential biases of archaeologists in the selection of attributes to be recorded.

Part 3 - Network Visualization

The third part of my thesis deals with a more theoretical problem from the field of network visualization: Can we draw a clustered network with overlapping clusters without crossings of edges and without “unnecessary” crossings of the borders of the clusters. This problem has been shown to be intractable (i.e. there are cases for which it can take an eternity to find a solution with current computer technology). We therefore focus on a subproblem, where the network induced by a cluster is connected. For this problem, my colleague Sabine Cornelsen and I were able to find an algorithm that calculates a solution within reasonable (i.e. polynomial) time. The results can be used to visualize networks such as the ones generated in the parts above including additional information such as the types of sherds found at a site, which can be encoded as a cluster.

By Jan Christoph Athenstädt