The anAIRlyze experiments improved the existing solution inventAIRy® that implements an indoor drone-based stock taking solution for warehouses, with a data analysis component. The generated raw data streams of the drone(s) may generally be separated into telemetry relevant data and observed product/object relevant data (Data-in-Motion). This data was visualized but not yet analyzed and processed. Those data streams in the experiment have been analyzed and transformed into higher level logistical KPI datasets (Data-in-Rest). The goal was to visualize the KPIs in a speciﬁc dashboard for the end user who is then enabled to take necessary countermeasures for process optimization.
During the anAIRlyze experiment conducted by doks. images of ﬂights and sensor values (e.g. from battery) were collected and saved to rosbag ﬁles for later oﬄine replaying and processing capability.
Upon further experimentation these datasets were used to ﬁnd empty positions with TensorFlow which gave good results. The playback of other sensor values helped in the battery prediction which was also tested with good results on additional ﬂights after the development and battery model creation.
To be able to compete, process quality must rise, so that human errors are identiﬁed and corrected on a regular basis. The approach of doks. within the warehouse is to have an automated drone solution doing this cross check. Humans still operate forklifts and store goods in pallet racks, pick products and pack them for goods out. The automated drone solution can ﬁnd those errors and report them for further investigation. The impact of such a solution is in general less errors been made, so that less search processes for speciﬁc objects are necessary. No more lost items which should be at a speciﬁc position but are somewhere else. With automated optical inspection and additional sensors as attached to such a solution even qualitative aspects may be affected.
34131 Kassel, Germania