DePILOT Reloaded

Important results from the project

DePILOT Reloaded consolidates the results of the original DePILOT project on the decoupling of pilots and drones. We explored emerging sensing technology, such as mmWave, to provide the drone with an additional level of perception. One specific example is accurate object detection, which is a key requirement in industrial settings (bit.ly/2VMk3EB). The work on designing novel drone interactions explored two directions, including the somatic explorations of how we could engage with drones and a collaboration with the fashion industry to find novel expressions for drones.

Expected long term effects

We built a custom drone using TI mmWave devices and developed a full-fledged signal processing pipeline that allows the drone to distinguish shape and material of an object within 4 meters. We tested the system in a total of 36 different configurations. We obtained 95%+ accuracy in the detection of objects within a 4cm error and a 98% accuracy in the detection of the corresponding material. The collaboration with the fashion industry is at the center of a research proposal for a demonstrator project with the Digital Futures center at KTH, submitted by RISE and SU.

Approach and implementation

The custom drone we built using mmWave technology works based on parameters and settings decided using machine learning technology. Based on a training set we created, the signal processing pipeline is dynamically adapted and allows the drone able to distinguish up to ten different materials, including plastic and water, for example. The design of novel drone interactions was entirely based on dedicated user studies, using knowledge and experience from reports obtained from professional drone pilots.