THE PROBLEM
Client, a leading manufacturer of industrial handheld mobile devices (such as barcode scanners), produces ruggedized scanning devices for usage in warehouses. Although these devices are ruggedized, they do face damages while in-use, which the client is obligated to repair under warranty claims. Hence, the client required a solution that would enable a deeper understanding of the device usage patterns in real life, especially when and how they get damaged which would enable the client to verify warranty eligibility, provide quicker maintenance, understand ruggedization requirements
INXITE OUT APPROACH
Data Capture Mechanism
On-device agent (customer Android application) was implemented to collect onboard accelerometer and gyro meter data to detect device movements.
Device Abuse Detection
- ML classification algorithm (Random Forest, SVM) were trained on accelerometer & gyro meter data to identify signatures of different events such as drop, impact, throws, slam (on a table or wall) and accidental drops
- False positives such as running with the device in the pocket or throwing the device followed by a catch were also handled
RESULT
Device abuse use cases were identified and classified from device usage data stream with ~90% accuracy