Migration from 20220927/0.5.1 to 20230114/0.7.1

The 20230114 release, which corresponds to Python SDK 0.7.1, brings a few breaking changes introduced to support the newest FW of the sensor, FW 3.0. The changes and how to mitigate them are summarized here.

Signal Multiplier

FW 3.0 allows setting the signal multiplier to non-int values 0.25 and 0.5. As such, we have changed the type of the sensor_config struct member signal_multiplier to double.

Old saved config jsons will continue to load properly.

XYZLut and sensor_info

The sensor_info struct now contains the beam_to_lidar_transform specifying the relationship between the beam coordinate frame and lidar coordinate frame. The double lidar_origin_to_beam_origin_mm which corresponds to the (0, 3) element of the beam_to_lidar_transform has not been removed.

The make_xyz_lut function now takes a mat4d transformation specifying the relation between the beam and lidar coordinate frames, as opposed to the previous double which assumed an Identity rotation.

Users using pre-FW3.0 sensors with the latest SDK make_xyz_lut overload which accepts a sensor_info need not change anything, as the sensor_info struct automatically derives and self-populates with the appropriate beam_to_lidar_transform.

Default Parameters in init_client

The shortform C++ init_client and Python client.Sensor() no longer have default parameters. Since they do not configure the sensor, it didn’t make sense to default to any value for any parameters. Users must provide the ports and hostname explicitly now.

Timeout Improvements

The Python Scans interface timeout parameter has been changed from None to 1 second by default to avoid confusing hanging behavior when timing out is appropriate. The default timeout has also been changed to 2 seconds across the board.

Notes for the future

For customers who know they will continue to upgrade their version of the SDK, we also wish to highlight some upcoming breaking changes.

Default Metadata Format

First, the next release will default writing the non-legacy metadata format as opposed to the current format, also known as the legacy metadata format. The SDK will continue to read the legacy format, i.e., it will continue to read old recorded data), and it will also be able to produce the legacy format if the parameter legacy=true is specified to the get_metadata function.

Deprecations

We are going to start removing a number of deprecations in the ouster_client code with the next release. We will document any such removals and how to migrate from them in this or the following migration guide (whichever is relevant).

  • LidarScan::N_FIELDS: The number of fields in a profile has varied since the introduction of eUDP profiles. To find the number of fields in your scan, we suggest using the iterator to loop over and count the number of fields.

  • LidarScan::Field: use sensor::ChanField instead.

  • LidarScan::BlockHeader: BlockHeaders, structs consisting of the timestamp, encoder ticks, and status which corresponded to a single column of measurements (identifiable by``measurement_id``), have been deprecated in favor of Headers, which are Eigen Arrays of length equivalent to the # of columns in a scan, with the ith element being the value from the ith measurement. As such, where once you might write:

    auto n_invalid = std::count_if( scans[i].headers.begin(), scans[i].headers.end(), [](const
    LidarScan::BlockHeader& h) { return h.status != 0xffffffff; });
    

you should now use the status() function:

auto status = scan.status(); auto n_invalid = std::count_if(status.data(), status.data() +
status.size(), [](const uint32_t s) { return !(s & 0x01); });

Timestamps are now available through timestamp(), also returning a Header; and the information contained in encoder is available through measurement_id() (see the next line item for the conversion).

  • encoder: Encoder counts were part of the LEGACY UDP profile, representing the azimuth angle as a raw encoder count starting from 0 to a max value of 90111. It thus incremented 44 ticks per azimuth angle in 2048 mode, 88 ticks in 1024 mode and 176 in 512 mode. To recover encoder infcount, you can multiply the measurement_id by the number dictated by your lidar mode. We would suggest, however, migrating to use simply measurement_id and multiplying by 360 degrees/ N where N is the number of columns in your mode (512, 1024, 2048) when you need the azimuth, thus untying any sense of azimuth from the internal mechanicals of the Ouster sensor.