Version
中文 | English

6.5. Performance Benchmark Tools

This section introduces the performance benchmark tools, their purpose, and the metrics they help measure.

6.5.1. common_benchmark_node.py

common_benchmark_node.py monitors Orbbec camera performance in a ROS environment. It collects and records subscriber-side metrics such as image frame rate, latency, system resource usage, and estimated frame loss to help evaluate camera node stability and performance. Statistics are updated once per second.

Features:

  • Measures the frame rate and latency of images received by the subscriber (current, minimum, maximum, and average).

  • Monitors camera node CPU/ARM usage (current, minimum, maximum, and average).

  • Estimates subscriber-side frame loss from image message timestamps.

  • Supports color, left_color, right_color, depth, ir, left_ir, and right_ir image streams, as well as compressed image and point cloud topics.

  • Prints real-time statistics at 1 Hz and saves results to a CSV file.

  • Supports configurable runtime, CSV output path, camera names, and topics.

The node estimates subscriber-side frame loss from adjacent image message timestamps. When --ideal_fps is set, it uses the specified ideal frame rate; otherwise, it learns the nominal frame interval from received image timestamps.

common_benchmark_ros1

Run with default settings:

rosrun orbbec_camera common_benchmark_node.py --run_time 30s

Run with a custom CSV output path:

rosrun orbbec_camera common_benchmark_node.py \
  --run_time 2h \
  --csv_file /path/to/log.csv

Multi-camera monitoring example:

rosrun orbbec_camera common_benchmark_node.py \
  --run_time 1h \
  --csv_file /tmp/cam_log.csv \
  --camera_names camera,camera01 \
  --topics color,left_color,right_color

Parameters:

  • --run_time / _run_time: Monitoring duration. Supported formats include 10s, 5m, 1h, and 2d. The default is 10s.

  • --csv_file / _csv_file: Output CSV file path.

  • --camera_names / _camera_names: Camera name list, separated by commas.

  • --ideal_fps / _ideal_fps: Ideal frame rate used for subscriber-side frame-loss estimation. A value greater than 0 overrides the interval learned from image timestamps.

  • --topics / _topics: Topics to monitor, separated by commas. Use raw stream names such as color,left_color,right_color, or full raw/compressed image and point cloud topics. Automatic discovery selects only raw image topics; compressed images and point clouds require full topic names.

6.5.2. service_benchmark_node.py

service_benchmark_node.py monitors service call performance. It calls multiple services from a YAML configuration file and reports call count, success rate, average time, minimum time, and maximum time.

Features:

  • Benchmarks multiple services defined in a YAML configuration file.

  • Builds requests automatically from the active ROS service type.

  • Can save benchmark results to a CSV file.

service benchmark

Multiple services benchmark (YAML configuration + CSV output):

rosrun orbbec_camera service_benchmark_node.py \
  --yaml_file /path/to/default_service.yaml \
  --csv_file /path/to/results.csv

The example YAML configuration is available at scripts/default_service.yaml. default_count is used when an entry does not specify its own call count.

6.5.3. service_benchmark_node

service_benchmark_node is the C++ service call benchmark tool. The Python version, service_benchmark_node.py, is recommended for most cases because it can batch-test services from YAML and write CSV output.

Single service benchmark:

rosrun orbbec_camera service_benchmark_node \
  _service_name:=/camera/set_color_ae_roi \
  _service_type:=orbbec_camera/SetArrays \
  _request_data:='{data_param: [0, 1279, 0, 719]}'

Multiple services benchmark (YAML configuration):

rosrun orbbec_camera service_benchmark_node \
  _yaml_file:=/path/to/default_service.yaml

6.5.4. Benchmark Recommendations

  1. Start with a small service set to verify network and device stability, then expand to the full service list.

  2. For long runs, write --csv_file to a persistent directory.

  3. Archive results by device and date for trend analysis.

  4. If abnormal latency appears, inspect the /camera/device_status topic together with the benchmark result.

For a complete service YAML example, see scripts/default_service.yaml in the source tree.