Fusion Sensor & Arduino Nano Ble Sense binaries not working

innomrk · March 18, 2022, 8:23pm

Hello,

I tried the new Fusion Sensor feature by combining all measurements from the Arduino’s IMU.

When I flash the binary(compiled from Edge Impulse) I get the following error when I try to run it on the board:

Failed to find sensor 'accX + accY + accZ + gryX + gryY + gryZ + magX + magY + magZ' in the sensor list

I also tried to use the Arduino library generated by Edge Impulse and chose the example nano_ble33_sense_accelerometer_continuous but I also get a similar error during compilation.

nano_ble33_sense_accelerometer_continuous:194:2: error: #error "Invalid model for current sensor"
 #error "Invalid model for current sensor"
  ^~~~~
exit status 1
#error "Invalid model for current sensor"

In the /src/model-parameters/model_metadata.h header file I have noticed this:

#define EI_CLASSIFIER_SENSOR                     EI_CLASSIFIER_SENSOR_FUSION
#define EI_CLASSIFIER_FUSION_AXES_STRING         "accX + accY + accZ + gryX + gryY + gryZ + magX + magY + magZ"

I guess the problem comes from the defined sensor type, which in the case is not EI_CLASSIFIER_SENSOR_ACCELEROMETER.

My questions are: Isn’t this supposed to work out of the box and if not, can somebody guide me on how to change the accelerometer example in order to get it working?

I think I need to format the sensor data as a string with the structure shown in EI_CLASSIFIER_FUSION_AXES_STRING but I am not sure.

rjames · March 19, 2022, 3:59pm

Hi @innomrk I’ll try to reproduce on my end

rjames · March 19, 2022, 4:38pm

@innomrk
The problem lies in your axes names. Fix the typo and change

gryX to gyrX
gryY to gyrY and
gryZ to gyrZ.

innomrk · March 19, 2022, 5:51pm

Thank you very much for pointing this out… I didn’t notice the typo at all… The interesting part is that the axes names were automatically generated by edge impulse.

rjames · March 19, 2022, 6:10pm

Can you share the steps you took to create the project? And more specifically how did you upload your data?

innomrk · March 19, 2022, 6:43pm

I used the edge-impulse-daemon on MacOS and I captured the data using the Arduino Nano BLE Sense. Check the attached screenshot. The axes are named wrong on data capture.

PS: The problem lies here:

github.com

edgeimpulse/firmware-arduino-nano-33-ble-sense/blob/7244f0dd0747976944c6497da5de297524af85ec/src/sensors/ei_inertialsensor.h#L51

      
        
            
            
float *ei_fusion_inertial_read_data(int n_samples);
            static const ei_device_fusion_sensor_t inertial_sensor = {
                // name of sensor module to be displayed in fusion list
                "Inertial",
                // number of sensor module axis
                INERTIAL_AXIS_SAMPLED,
                // sampling frequencies
                { 20.0f, 62.5f, 100.0f },
                // axis name and units payload (must be same order as read in)
                { {"accX", "m/s2"}, {"accY", "m/s2"}, {"accZ", "m/s2"}, {"gryX", "m/s2"}, {"gryY", "m/s2"}, {"gryZ", "m/s2"}, {"magX", "uT"}, {"magY", "uT"}, {"magZ", "uT"} }, 
                // reference to read data function
                &ei_fusion_inertial_read_data
            };
            
            
#endif

rjames · March 19, 2022, 6:53pm

Thanks for the response. Yea, we’ve fixed it upstream but haven’t (yet) pushed the fix to this repo. Thanks for pointing that out. I’ll provide a hot-fix.

rjames · March 19, 2022, 7:03pm

The repo is updated now!

innomrk · March 19, 2022, 7:13pm

Perfect! Thanks for the help!