Gyroscope/compass Linux/Jetson Nano driver

[ntwong0]

progress

• [x] get raw data from MPU-9250
• [x] Decode compass data
• [x] read data with ros-mpu9250-node
• [ ] Determine compass calibration procedure

useful links

• ros node (dependency doesn't support jetson nano on 10/05/2020)
• magnetometers data to compass heading
• MPU9250 Datasheet
• MPU9250 Register Map

ref for i2c timing diagram

[ntwong0]

@BobMak was not able to look at compass node this week

[ntwong0]

[BobMak]

Node code appears to be calling the constructor with a correct bus specified. The problem appears to be on mraa side - some initialization failure.

[BobMak]

trying to set raw=TRUE in m_i2c(int bus, boolean raw)
https://github.com/eclipse/upm/blob/master/src/mpu9150/ak8975.cxx#L38

[ntwong0]

Looking into mraa/i2c.h, we should try to use mraa_i2c_init_raw(), so this is a good approach.

@BobMak mpu60x0.cxx also needs to be modified at line 36, since this is the inherited constructor for MPU9250.

[ntwong0]

No updates for this week, will check back on @BobMak next week

[BobMak]

errors in making upm from the source. Need to resolve dependencies. so far tried apt-get install libbacnet libmodbus libopenzwave libopenzwave

[ntwong0]

@BobMak can you attempt building upm from source by next Saturday? Stretch goal - draft of calibration procedure

[ntwong0]

@BobMak @YMmirsky need clarification on next steps

[ntwong0]

@BobMak still blocked on upm building, will meet with @YMmirsky to resolve

[BobMak]

Able to build upm. upm requires a specific nodejs version to build optional utilities. It is possible to build upm without node by running cmake <upm_directory> DBUILDSWIGNODE=OFF Original solution

[PushpalPatil]

@BobMak Note: saturday

[BobMak]

Rebuilding upm with raw=TRUE results in the same error

[BobMak]

Core issue. Replacing mraa i2c read/write calls in upm.

[BobMak]

I've replaced the mraa i2c calls with linux i2c code in the upm as a temporary workaround. I can see the accelerometer data, but the orientation data looks broken.

[PushpalPatil]

maybe use magnet method

[BobMak]

Tried enabling magnetometer by setting the enableAk8975 flag to true. Still getting garbage data, no noticeable changes when the magnet is put closer.

[PushpalPatil]

can you post an example or vid recording of the garbage data? also could you push your code to the repo so that we can see the code?

[BobMak]

garbage data and magnet test

[BobMak]

Failed to utilize the ros-mpu9250-node module on Jetson Nano. The low-level dependency of the node doesn't support Jetson nano yet. Replacing the dependency failed. Working on my own magnetometer node implementation. Already able to initialize and read from the magnetometer.

[BobMak]

working on a python mpu9250 orientation node implementation.

[BobMak]

vizualised pose data by using Nelson's rviz notes there was one other thing I needed to run to see the live data visualisation: rosrun tf static_transform_publisher 0.0 0.0 0.0 0.0 0.0 0.0 map imu 100

mpu orientation data looks consistent and is reacting to a magnet. now trying to figure out how to do self-test and how to calibrate the sensor.

[BobMak]

mpu9250 compass heading visualization