pom-genom3 3.2 released

Added by Anthony Mallet 6 months ago


  • Handle magnetometer in the measurement sources. Magnetometers must publish the magnetic field in the attitude part of the state, as an intrinsic measurement, using the vector part of the quaternion and a NaN scalar part.
  • Add set_mag() service, setting the expected magnetic field of the current location, in the NWU (X north, Y west, Z up) world frame.



pom-genom3 3.1 released

Added by Anthony Mallet about 1 year ago


  • Add an attribute selecting the prediction model (constant velocity or constant acceleration)
  • Fix the replay service so that it can read newer log files
  • Add the variance of the state in the state log file
  • Log the delay of measurements (current time versus measurement time stamp)



pom-genom3 3.0 released

Added by Anthony Mallet over 1 year ago


  • Switch to openrobots2-idl
  • Fix syntactic issues with older eigen
  • Display the covariance matrix diagonal when it becomes not positive definite
  • Allow non-integer history length (to make it less than one second)
  • Warn if computation is too slow (more than 110% of the period for 2 periods in a row)



pom-genom3 2.3 released

Added by Anthony Mallet over 2 years ago


  • Fix conversion from euler angles to quaternion in the 'replay' service * Abort processing of history after computing for longer than the period. This helps in keeping up the desired output rate when many old measurements are received. * Use aio(7) for logging and drop log entries when io requests are too slow * Add a 'decimation' parameter to the log_state function * Decouple the position and orientation measurements. This allows a measurement producer to provide only the translation part of a pose (e.g. GPS, ...) or the orientation part (e.g. compas). * Try to fix non positive definite covariance matrix in the state. Altough this does not happen in practice, it can still be triggered when receiving very wrong measurements. * Add README.adoc from latest genom skeleton



pom-genom3 2.1 released

Added by Anthony Mallet about 4 years ago


  • Implement a realistic process noise, with in particular all cross correlations * Reject measurements older than the current history (2s by default). * Add 'replay' and 'log_measurement' services to record/replay raw measurements * Be robust to port disconnection when reading measurement data * Export covariance on velocity and acceleration




    Also available in: Atom