Cell Multi-Bernoulli (Cell-MB) Sensor Control for Multi-Object Search-While-Tracking (SWT).

Area of Science:

• Computer Vision
• Robotics
• Sensor Fusion

Background:

• Intelligent sensors optimize measurements using environmental feedback.
• Information-driven control in object tracking relies on information gain to reduce uncertainty.
• Random Finite Set (RFS) theory offers a framework for information gain estimation in multi-object tracking.
• Estimating information gain in RFS remains computationally intensive.

Purpose of the Study:

• To present a computationally tractable approximation of RFS expected information gain.
• To enable effective sensor control for multi-object search and tracking applications.
• To account for non-ideal measurement conditions in information gain estimation.

Main Methods:

• Developed a novel approximation for RFS expected information gain.
• Incorporated contributions from noisy measurements, missed detections, false alarms, and object state changes.
• Applied the approximation to sensor control algorithms for multi-object tracking.

Main Results:

• The proposed approximation is tractable and applicable to sensor control.
• The method effectively handles non-ideal measurement scenarios.
• Demonstrated improved sensor control in search-and-track experiments.

Conclusions:

• The new information gain approximation facilitates intelligent sensor control for multi-object tracking.
• This approach offers a practical solution to computationally challenging RFS estimation problems.
• Validated through real-world multi-vehicle search-and-track experiments using diverse sensor data.