voxl 2 rebooting when tflite server is ran

[Jetson Nano]

@Alex-Kushleyev @Vinny @Eric-Katzfey @Zachary-Lowell-0

Hey , I am facing an unusual error. where when i try to start tflite server, the voxl 2 board is rebooting up. it is only when i run the server or try to visualize it in the portal.
I tried with the latest tflite version still the same error.

could anyone help me with this

voxl2:/$ voxl-tflite-server 
=================================================================
skip_n_frames:                    0
=================================================================
model:                            /usr/bin/dnn/lite-model_movenet_singlepose_lightning_tflite_float16_4.tflite
=================================================================
input_pipe:                       /run/mpa/hires_front_small_color/
=================================================================
delegate:                         gpu
=================================================================
allow_multiple:                   false
=================================================================
output_pipe_prefix:               mobilenet
=================================================================
existing instance of voxl-tflite-server found, attempting to stop it
INFO: Created TensorFlow Lite delegate for GPU.
ERROR: Following operations are not supported by GPU delegate:
ARG_MAX: Operation is not supported.
CAST: Not supported cast case
CONCATENATION: OP is supported, but tensor type/shape isn't compatible.
DEQUANTIZE: 
FLOOR_DIV: OP is supported, but tensor type/shape isn't compatible.
GATHER_ND: Operation is not supported.
MUL: OP is supported, but tensor type/shape isn't compatible.
PACK: OP is supported, but tensor type/shape isn't compatible.
RESHAPE: OP is supported, but tensor type/shape isn't compatible.
SUB: OP is supported, but tensor type/shape isn't compatible.
UNPACK: Operation is not supported.
97 operations will run on the GPU, and the remaining 200 operations will run on the CPU.
INFO: Initialized OpenCL-based API.
INFO: Created 1 GPU delegate kernels.
INFO: Created TensorFlow Lite XNNPACK delegate for CPU.
Successfully built interpreter

------VOXL TFLite Server------

 4 5 6
 4 5 6
Connected to camera server
^C
received SIGINT Ctrl-C

[Alex Kushleyev]

Can you please clarify at which exact point does the reboot happen? Also, how are you powering the VOXL2 ? Can you try to replace the 4-pin Molex power cable (between VOXL2 and APM / ESC, whichever is providing power) and see if the issue persists?

Alex

[Jetson Nano]

@Alex-Kushleyev
the voxl 2 boots, when i click on the voxl-portal to see the tflite detection,that point of time the board reboots.
at the same time i am getting the hires camera data. and all other camera feeds.
this is the first time i am facing this kind of error.

I dont think it is a cable issue, since everything else is working. The same time, I am using VOXL 2 to FC v2 to ESC.

[Alex Kushleyev]

@Jetson-Nano , when you view tflite output from voxl-portal, the tflite processing actually starts, which causes a surge in cpu and gpu load, which causes an increased power draw. If the power cable has worn contacts (with increased resistance), the voxl2 board can go in undervoltage condition and reboot. With this in mind, could you please try to replace the power cable and let us know if this solves your issue?

Thank you

Alex

[Jetson Nano]

@Alex-Kushleyev Thank you for reverting back, When i tested again without replacing the cable the board was not rebooting. Either way I will replace the cable and get back to you.

could anything else cause this error

[Alex Kushleyev]

@Jetson-Nano , generally it is not possible to cause a VOXL2 reboot during normal operation. There are a few cases which can cause a kernel panic and the board would reboot into a recovery mode for debugging. This only occurs during testing some experimental features and you would normally not see this behavior. If the board simply reboots and is running again after reboot, it is very likely a power failure.

The power cable, if wiggled a bit, can become temporarily normal again, so it is possible what happened to you. The main issue is that wiggling the power cable, in general, loosens the contacts that make contact with the pins, causing the contact pressure to weaken and contact resistance to increase.

Alex

[Jetson Nano]

@Alex-Kushleyev Thank you for reverting back.