Multi-Camera Multi-Sensor SLAM

Problem

Conventional single-camera SLAM systems often degrade under occlusion, motion blur, limited field of view, and sensor-specific failure modes. A more robust system needs to combine complementary observations across multiple cameras and additional sensors.

Method

This project studies a multi-camera, multi-sensor SLAM pipeline that jointly uses overlapping and non-overlapping cameras together with inertial and auxiliary sensing for state estimation, tracking, and map building. The emphasis is on cross-sensor calibration, synchronized fusion, and consistent optimization across heterogeneous measurements.

Focus

• Multi-camera geometry and calibration
• Visual-inertial and multi-sensor fusion
• Robust tracking under fast motion and partial sensor failure
• Consistent mapping across wider fields of view