Using the Robot Trace tool to debug Stretch Body

Starting with Stretch Body 0.4.26 and Stretch Factory v0.3.15 we have introduced a new debugging tool for Stretch Body called Robot Trace.

Note: This tool is recommended for advanced users only

The Robot Trace tool periodically logs Stretch Body status data to disk in the background. This data can then be inspected and plotted offline. The Trace tool can be useful to capture and analyze events when debugging transient issues and behavior with the robot.

For example, if the robot powers down unexpectedly during an experiment, you might look at the trace and plot the battery voltage and current at the time of the shutdown.

The trace function is turned off by default. To turn it on, update your stretch_user_params.yaml with:

  use_trace: 1

Now, whenever an instance of the Robot class is created a thread will be spawned that logs data to disk. The data is timestamped and stored under ~/stretch_user/log/trace. The amount of trace data logged to disk is bounded in a FIFO type queue.

ls ~/stretch_user/log/trace/

By default the robot status data that is logged is defined in The YAML trace files contain a series of entries like this:

  arm.motor.effort_pct: 0.0
  arm.pos: -0.019642551395004437
  arm.vel: 0.000298045866133731
  base.left_wheel.effort_pct: 0.0
  base.left_wheel.pos: 0.15481062233448029
  base.left_wheel.vel: 0.018910381942987442
  base.right_wheel.effort_pct: 0.0
  base.right_wheel.pos: 0.3108431398868561
  base.right_wheel.vel: 0.0069570220075547695
  lift.motor.effort_pct: 12.49999116448795
  lift.pos: 0.5268116892742208
  lift.vel: 0.000220643355869542
  pimu.charger_connected: false
  pimu.cpu_temp: 40.0
  pimu.current: 2.5582551079637863
  pimu.low_voltage_alert: false
  pimu.runstop_event: true
  pimu.voltage: 12.420293092727661
  timestamp: 1678299646.3529084

You can add your own call back function if you’d like to trace additional data.

Run the tool `’ to plot the data. You can see here that there is a menu to let you select a sensor value and time series to plot:

Plotting looks like:

Happy debugging!