This is part 4 of the new award winning development IoT kit from Nordic NRF54LM20-DK Nordic Semiconductor ASA | Development Boards, Kits, Programmers | DigiKey
Please refer to the previous articles Part 2 and Part 3 in order to properly setup this classification demo. This AXON NPU classification demo is a multi-class classifier that identifies parcel delivery states:
- Idle
- Shaking
- Impact
- Free Fall (Please see this previous article to learn more about a traditional approach for parachute release system)
- Carrying
- In Car
- Placed
from a stream of acceleration magnitude samples in the delivery process of the package. The potential benefits related to this application are but not limited to the following:
- Detecting rough handling or potential damage during shipment process
- Monitoring the delivery quality and other logistics situations
- Alerting when packages go thru impacts or free fall conditions
- Tracking parcel state transitions throughout delivery process lifecycle
Briefly we will cover here the Idle state, which means rest state (other states are defined as needed per requirements) but will not be covered to keep this article as brief as possible,
Characteristics of the Idle state:
- Acceleration values cluster tightly around 1000 mG (gravitational acceleration baseline)
- Minimal variation between consecutive samples (stable, static state)
- No significant spikes or sudden changes in acceleration
- Represents a parcel at rest on a surface (shelf, table, ground, stationary truck)
Use cases:
- Parcel stored at warehouse awaiting shipment
- Parcel on delivery truck stationary between stops
- Parcel at customer delivery location
- Parcel placed on shelf or table
These data characteristics are consistent baseline with minimal noise, typical for non-relativistic Newtonian simple gravitational field as shown in the next representative plot,
In order to compile this parcel delivery classification state demo proceed as follows,
~/DigiKey_Coffee_Cup/Nordic/$ source venv/bin/activate
then open the required shell,
(venv)~/DigiKey_Coffee_Cup/Nordic/$ nrfutil sdk-manager toolchain launch --ncs-version v3.3.0-preview2 --shell
Initializing shell environment!
(v3.3.0-preview2)
then proceed to build the application,
(v3.3.0-preview2) ~/DigiKey_Coffee_Cup/Nordic/$ west build -p always -b nrf54lm20dk/nrf54lm20b/cpuapp edge-ai/samples/nrf_edgeai/classification/
...
[260/260] Linking C executable zephyr/zephyr.elf
Memory region Used Size Region Size %age Used
FLASH: 56416 B 1940 KB 2.84%
RAM: 8720 B 511 KB 1.67%
IDT_LIST: 0 GB 32 KB 0.00%
...
Finally connect the USB cable to the the new award winning development IoT kit from Nordic NRF54LM20-DK Nordic Semiconductor ASA | Development Boards, Kits, Programmers | DigiKey shown below
Open a minicom terminal and then flash the application,
(v3.3.0-preview2) ~/DigiKey_Coffee_Cup/Nordic/$ west flash
In the minicom terminal, the Axon NPU classification demo is displayed,
It initializes the neural network, executes INFERENCE on 7 representative test cases covering all parcel states, and validates classification accuracy. Each time a complete 50-sample accelerometer window is processed, the model outputs its prediction and probabilities for each possible classification. The model does not require additional context beyond each window to make its predictions. The predicted class for each window is the one with the highest confidence probability.
This has completed the parcel package classification demo for this new Axon Neural Processing Unit (NPU) using the Axon NPU driver directly, in this innovative Nordic IoT platform. Please stay tuned for our next part in this series. This new Embedded World 2026 award winning Nordic IoT NRF54LM20-DK development kit is available at DigiKey.
Have a great day!
This article is available in spanish here.
Este artículo está disponible en español aquí.