This package includes ROS driver software for the InvenSense MPU9250 9DoF IMU. A ROS node publishes data from the IMU, and permits configuration through the ROS Parameter Server. A Raspberry Pi driver is included, but users may also easily create other device-specific drivers by extending the base driver class.

License: MIT

Author: Paul D'Angio

1. Installation: Instructions for installing this ROS package.
2. Usage: How to configure and use the driver_mpu9250 ROS node.
3. Extending: How to easily add your own custom device driver.
4. Additional Information: Other important information relevant to this package.

• Robot Operating System (ROS) (middleware for robotics)
• sensor_msgs_ext (ROS extended sensor messages)
• pigpio (Raspberry PI I/O)

To build from source, clone the latest version from this repository into your catkin workspace and compile the package using

# Switch to your catkin workspace directory.
cd catkin_workspace

# Clone the package into the workspace src directory.
git clone https://github.com/pcdangio/ros-driver_mpu9250.git src/driver_mpu9250

# Build the package.
catkin_make

This package includes a ROS node that publishes data from the IMU and permits configuration through the ROS Parameter Server.

The table below describes parameters may be set on the ROS Parameter Server to configure the MPU9250. These parameters must be set before the ROS node is run. Recall that ~ is the node's private namespace.

Run any of the driver nodes with (where xxx is the driver type):

rosrun driver_mpu9250 driver_mpu9250_xxx

To use the driver node for a Raspberry Pi:

# Make sure that the linux user has been added to the "gpio" group.

# In one terminal, start up the PiGPIO Daemon.
sudo pigpiod

# In a second terminal, start up the driver.
rosrun driver_mpu9250 driver_mpu9250_rpi

This package uses polymorphism with a base driver class to enable drivers to be easily built for a variety of hardware interfaces. To create your own device-specific driver:

• Create a new class for your device driver that extends the base driver class in src/driver.h
• In your derived driver class, implement the several pure abstract functions of the base driver class
• Create a new main.cpp that instantiates your derived class and passes it to the src/ros_node.h instance through polymorphism before running the node.
• Add the new main.cpp to your CMakeLists.txt as a new node executable.

You can refer to the following files in this package as an example for creating your custom device driver:

src/rpi_driver.h
src/rpi_driver.cpp
src/main_rpi.cpp
CMakeLists.txt

If you have created and tested a driver for a device not already covered in this package, please feel free to submit a pull request!

The "full-scale range", or FSR, of a sensor specifies the range of measurements that must be sensed. Smaller ranges will measure at higher resolution, while larger ranges will measure at lower resolution. The MPU9250 has several full-scale ranges to select from, as shown in the table below. The Enumerated Value is the integer that should be used in the ~/gyro_fsr and ~/accel_fsr ROS parameters during configuration.

The MPU9250 has internal Digital Low-Pass Filters (DLPF) that can be used to filter out unwanted noise from the gyroscope, accelerometer, and temperature measurements. The table below lists the available cutoff frequencies that can be chosen. Lower cutoff frequencies will filter out more noise, but will increase the lag (delay) of the measurements. The cutoff frequency selected for the gyroscopes is also applied to the temperature sensor. The Enumerated Value is the integer that should be used in the ~/gyro_dlpf_frequency and ~/accel_dlpf_frequency ROS parameters during configuration.

The data rate is the frequency at which the sensor makes measurements and publishes them through ROS. This driver automatically calculates the appropriate data rate based on the cutoff frequencies of the DLPFs of the acceleremeters and gyroscopes. The driver finds which DLPF has the highest cutoff frequency, and multiplies that frequency by 2.5 to calculate a data rate with a proper Nyquist sampling rate. The calculated data rate is output to the node's ROS log at startup.

The ~/max_data_rate parameter can be used to override the calculated data rate to a lower frequency.

Accelerometers, gyroscopes, and magnetometers all require calibration to provide accurate measurements:

• Accelerometers suffer from axis misalignment and bias.
• Gyroscopes suffer from bias and bias drift.
• Magnetometers suffer from hard iron and soft iron distortion.

This driver handles calibrations differently for each sensor:

The MPU9250's 3-axis accelerometer and 3-axis magnetometer have relatively stable bias (assuming no changes in sensor mounting and metals/magnets surrounding the magnetometer). Thus, the driver uses static calibration matrices to make adjustments to sensor data before it is published through ROS. These matrices can be calculated using the calibration_imu package, and set as parameters for the driver to load and use. NOTE: You should recalibrate the accelerometer and magnetometer any time you make hardware changes to your platform.

The MPU9250's 3-axis gyroscope, like all MEMs gyroscopes, suffers from continuous bias drift. Thus, it is necessary to regularly recalibrate the gyroscope when it is known that the IMU is not in motion. This calibration routine can be called through a service, which calculates the current bias of the gyroscope and subtracts it from future measurements.