Archaeological excavations typically last several weeks and may unearth thousands of individual pottery sherds. For logistical and legal reasons, archaeologists often cannot bring these materials back from the field and are only able to return home with the data that they collect about these sherds. This makes careful data collection an integral part of archaeological fieldwork. For example, high resolution photos can capture color, shape, and size information of a sherd and this, along with measurements of the mass of each sherd, can be used to determine the material composition of the artifact. However, taking thousands of measurements is a dull and repetitive task that requires a great deal of time to properly accomplish. Furthermore, the time spent cataloging data about thousands of small, unremarkable body sherds (i.e. pieces which do not have any distinguishing geometric information) is time that is not spent digging, studying the more informative diagnostic sherds (pieces of handles, rims, bases, etc.), or performing higher-level cognitive tasks, such as in situ interpretation of the current findings to help inform where to dig in the following days.

This project will develop and demonstrate the feasibility of using a robotic arm to autonomously process ceramic sherds during archaeological fieldwork. Specifically, we will use state-of-the-art machine learning and computer vision algorithms to create a suite of open-source software tools to: 1) reliably locate, identify, pick up, and transport sherds and 2) interface with external sensors to collect high quality data about each sherd (e.g. mass, color, shape). While this project will not completely obviate the need to manually catalog sherds in the field, it will offload a significant fraction of the labor from humans to the robotic system. Note that while this proposal uses ceramic sherds as its model artifacts, the proposed robotic arm could also be used to handle other types of archaeological artifacts. Also note that we are not proposing to use the arm to unearth artifacts, but rather to create a detailed digital record of artifacts that can be used, along with the context of where the artifacts were found, for future analysis.

Publications

• Peter Cobb, Tiffany Earley-Spadoni, Philip Dames.  “Centimeter-Level Recording for All: Field Experimentation with New, Affordable Geolocation Technology” Advances in Archaeological Practice.
  [DOI] (Open Access)