Stretch-ICP: A Continuous-Trajectory Registration and Deskewing Algorithm in Scenarios of Aggressive Motions

Area of Science:

• Robotics
• Computer Vision
• Sensor Fusion

Background:

• Robotic autonomy in complex terrains faces challenges due to instability, leading to extreme motion that corrupts sensor data.
• Degraded sensor measurements and state estimation hinder robust robotic performance in real-world scenarios.

Purpose of the Study:

• To address the challenge of robust state estimation in aggressive robotic motions.
• To introduce a new dataset and algorithms for improving robotic localization and mapping.

Main Methods:

• Introduced the Tumbling-Induced Gyroscope Saturation (TIGS) dataset with high angular speeds.
• Developed Saturation-Aware Angular Velocity Estimation (SAAVE) to handle gyroscope saturation.
• Proposed Stretch-ICP, a novel algorithm for improved registration and trajectory reconstruction.

Main Results:

• SAAVE reduced angular speed estimation error by 83.4% during aggressive motions.
• Stretch-ICP reduced linear and angular velocity errors by 95.2% and 94.8% at scan boundaries.
• The proposed methods enhance the robustness and consistency of lidar-inertial state estimation.

Conclusions:

• The TIGS dataset and proposed algorithms significantly advance lidar-inertial state estimation robustness.
• These contributions enable more reliable robotic autonomy in challenging, dynamic environments.