As I write this, Embedded World 2024 has just finished in Nuremberg, Germany. This year’s show was highly attended and huge in scale as usual. The interest level for new sensor use cases and AI/ML was strong as we had a steady stream of visitors to our booth asking questions and exchanging application needs for edge IoT AI/ML. It’s always interesting and insightful to talk with prospective new customers and partners and understand their needs. Those we spoke to represented a broad range of applications from agricultural sensing to industrial anomaly detection and included IoT device OEMs, systems integrators, silicon vendors, startups, and universities.
In the SensiML Booth
In the SensiML booth, we led with our always popular boxing game which was front-and-center in the booth with a fun giveaway contest that showcased a gesture recognition wearable use case built with the SensiML Toolkit. It was entertaining to watch all the different punching styles people displayed from downright tame to at one point some concern about whether our monitor would end up flying off the demo pod! Fortunately, no electronics ended up injured over the three days, though after the fact, I wished we had captured all those various styles we saw in the Data Studio Boxing Game project. Maybe next time….
We also showcased a proof-of-concept for a smart drill that uses SensiML AI/ML edge models to transform power tool functionality. By utilizing motion sensors (6-axis accel/gyro data), drill motor voltage, and variable speed trigger input voltage, we showed how we can classify the various states associated with driving a screw fastener with the drill.
The application concept is to replace the mechanical clutch found on most handheld power drills with motor control circuitry driven by real-time edge classification. By recognizing the user’s drilling task and then the state of the drill within that task, the passive physical clutch can be replaced by a smart-sensing ML algorithm that automatically controls the motor to achieve the desired task faster, safer, and with less workpiece damage. The booth demo classified the task of driving screw fasteners with real-time state detection of driving and removing screws into a wood block along with transient states for bit slippage (i.e. cam out) and fully driven fastener. The resulting model achieved >98% accuracy with an ML model that requires only 5.5kB of combined flash and SRAM to execute.
SensiML Elsewhere at Embedded World
Both Arrow Electronics and Arduino Pro featured the SensiML predictive maintenance use case in their booths using variations on the SensiML Fan Demo.
The Arrow demo showed a working edge ML model for fan state operation running on Silicon Labs MG24 board, while Arduino’s booth showed yet another model supported on the Arduino Nicla Sense ME.
Speaking of our partner Silicon Labs, they also helped promote the smart drill proof-of-concept with their own static display of an MG24 equipped drill and sent interested attendees down to view the demo in detail at our booth. Thanks to Silicon Labs for the coordination on showing this use case!
If you missed us at Embedded World this year, please come and see us at an upcoming event. SensiML will be exhibiting at:
- Tiny ML Summit
April 22- 24 Burlingame, CA - Sensors Converge
June 24-26, Santa Clara, CA
