To address the reliability issues of collaborative utilization of multi-source data in intelligent transportation environmental perception

a multi-object tracking algorithm based on a real-time modular multi-sensor fusion framework is proposed and verified. This method aims to fuse object list-level data from distributed cameras

radars

and LiDARs

and to construct a tracking architecture by integrating coordinate transformation

object association based on the Hungarian algorithm

unscented Kalman filter-based object tracking

and motion compensation. Results show that the fusion framework

tested using simulation and public domain experimental data

meets real-time requirements. Its modular design is independent of the specific number or type of sensors and can continue operating even when single sensor fails

enhancing system robustness. Furthermore

this method can detect changes in external sensor calibration and potential faults. In conclusion

this adaptive framework provides reliable environmental perception capabilities for safety-critical applications

and its built-in fault identification concept is of great significance for ensuring the functional safety of autonomous driving.