Core Concepts
A tightly-coupled LiDAR-Inertial-Kinematic Odometry (LIKO) system that achieves high-frequency and accurate state estimation for bipedal robots by leveraging LiDAR, inertial, and kinematic measurements.
Abstract
The paper presents a tightly-coupled LiDAR-Inertial-Kinematic Odometry (LIKO) system for accurate state estimation of bipedal robots. The key highlights are:
LIKO uses an iterated extended Kalman filter to tightly fuse LiDAR, inertial, and kinematic measurements for high-frequency (1kHz) and globally accurate state estimation.
It models and estimates the foot contact position, allowing for both position and velocity updates from the kinematic measurements. This improves the velocity estimation compared to previous methods.
The authors collected a dataset of a bipedal robot (BHR-B3) with LiDAR, IMU, joint encoders, force/torque sensors, and motion capture ground truth. Experiments show LIKO outperforms state-of-the-art LiDAR-inertial and inertial-kinematic odometry methods by 14% in accuracy.
The ablation study demonstrates the individual contributions of different sensor modalities. LiDAR is crucial for global position estimation, while the high-frequency kinematic measurements provide accurate velocity updates.
LIKO's accurate state estimation, including contact position, can benefit downstream control applications for more robust bipedal robot locomotion.
Stats
The robot's linear velocity in the X direction reaches up to 0.5 m/s.
The robot's linear velocity in the Y direction reaches up to 0.5 m/s.
The robot's linear velocity in the Z direction reaches up to 0.1 m/s.
Quotes
"High-frequency and accurate state estimation is crucial for biped robots."
"Beyond state estimation, the foot contact position is also modeled and estimated. This allows for both position and velocity updates from kinematic measurement."
"The dataset and source code will be available at https://github.com/Mr-Zqr/LIKO."