3D LiDAR + IMU SLAM Navigation
Tightly coupled factor-graph SLAM combining 64-beam LiDAR point clouds with 400 Hz IMU pre-integration for sub-centimeter indoor localization.
ros2 launch ckw_amr lio_sam.launch.py \
voxel_size:=0.05 imu_rate:=400
Open technical blueprints, 3D LiDAR SLAM sensor fusion pipelines, quantized TensorRT vision benchmarks, and deterministic Linux kernel tuning for robotic manipulators and autonomous mobile platforms.
Modular ROS 2 Humble nodes for spatial mapping, neural depth estimation, and deterministic actuator control.
Tightly coupled factor-graph SLAM combining 64-beam LiDAR point clouds with 400 Hz IMU pre-integration for sub-centimeter indoor localization.
ros2 launch ckw_amr lio_sam.launch.py \
voxel_size:=0.05 imu_rate:=400
Hardware-accelerated INT8/FP16 object detection and semantic occupancy grids running with zero-copy NITROS buffers on embedded GPUs.
trtexec --onnx=yolov8m_seg.onnx \
--fp16 --saveEngine=vision.engine
CPU core isolation (`isolcpus=2,3`), IRQ pinning, and lockless ring buffers for hard real-time robotic joint trajectory controllers.
sudo cyclictest --smp -p 95 \
-i 200 -d 0 -D 10m
Standardized hardware schematics for quasi-direct-drive (QDD) quadruped actuators and industrial cobot arms.
Integrated 18-bit magnetic off-axis encoder, FOC field-oriented controller, and 9:1 planetary reduction for legged locomotion.
Zero-jitter IgH EtherCAT master stack with Distributed Clocks (DC) synchronization across 12 robotic joints.
Adjust payload mass, reach length, and angular acceleration to compute joint torque and visualize manipulator posture.