What is needed
Measure FusionCore CPU usage on Jetson Orin Nano (8GB) at 100 Hz with a realistic sensor load:
- IMU at 100 Hz
- Wheel encoder at 50 Hz
- GPS at 5 Hz
- Optional: second IMU at 100 Hz (imu2.topic)
Report:
- Mean and peak CPU time per filter cycle (µs)
- RAM usage
- CPU core utilization
- Comparison: with and without
imu2.topic enabled
Why it matters
Evaluators comparing FusionCore against other stacks (TIER IV EagleEye, Wolf, robot_localization) want a concrete number before committing to a platform. The Raspberry Pi 4 figure (under 1 ms per cycle) is documented, but Jetson is the more common deployment target for F1/10 and van platforms.
Method
# Build with Release flags on Jetson
colcon build --cmake-args -DCMAKE_BUILD_TYPE=Release
# Profile with perf or ros2 topic hz + /proc/pid/stat
# Use a rosbag replay at real-time speed with a fixed sensor load
Expected result
Based on the Pi 4 baseline and Jetson Orin's Cortex-A78AE cores, expected under 0.5 ms per cycle. Needs verification.
What is needed
Measure FusionCore CPU usage on Jetson Orin Nano (8GB) at 100 Hz with a realistic sensor load:
Report:
imu2.topicenabledWhy it matters
Evaluators comparing FusionCore against other stacks (TIER IV EagleEye, Wolf, robot_localization) want a concrete number before committing to a platform. The Raspberry Pi 4 figure (under 1 ms per cycle) is documented, but Jetson is the more common deployment target for F1/10 and van platforms.
Method
Expected result
Based on the Pi 4 baseline and Jetson Orin's Cortex-A78AE cores, expected under 0.5 ms per cycle. Needs verification.