Anticipate weak observations
Ray tracing supports predictions of the information available to the sensor.
IEEE RA-L · 2025 / PERCEPTION & CONTROL
UA-MPC.
Move the sensor where the information is.
Uncertainty-aware control that adapts a motorized LiDAR’s scanning motion to the surrounding geometry.
- LiDAR odometry
- Model predictive control
- Active sensing
01 / OVERVIEW
The idea.
UA-MPC: Uncertainty-Aware Model Predictive Control for Motorized LiDAR Odometry
IEEE Robotics and Automation Letters · 2025
A motorized LiDAR can change its scanning direction. UA-MPC uses predicted observability to choose that motion, balancing sensing accuracy and efficiency as the system encounters different environments.
- Uncertainty-aware motor control
- Geometry-dependent scanning
- Real-world 3D mapping
RESEARCH VISUALS
Watch the experiment.
A short preview from the coauthor’s UA-MPC materials. The original link opens the NTU Hive experiment.
02 / IMPLEMENTATION
How it works.
- Scan
Observe the nearby environment.
- Predict
Estimate future observability from geometry.
- Optimize
Select motor motion with predictive control.
- Map
Use the resulting scans for odometry.
Adapt the scanning motion
Model predictive control adjusts the motor instead of fixing its speed.
Map challenging spaces
The demonstrations include the NTU Hive and Spine environments.
03 / RESOURCES
Read further.
The paper, original demonstrations, and available project materials.