Environmental Perception and Sensing Systems for Intelligent Control: Multi-sensor Fusion and Closed-Loop Control Methods

This paper reviews how perception systems in intelligent systems evolve from single sensing to multi-source collaboration and then to participation in closed-loop control. Sensors such as vision, LiDAR, radar, and IMU differ in spatial coverage, accuracy, and environmental adaptability, making fusion necessary; time synchronization and spatial alignment are its prerequisites. Each of the data-level, feature-level and decision-level fusion is appropriate to certain tasks. Calibration of sensors, degradation detection, and fault detection in complex conditions guarantee long-term reliability, and machine learning enhances the extraction of information. Control decisions are becoming more and more driven by perception information via state feedback such that quadruped robots are using multi-IMU proprioception to drive MPC switching and connected vehicles are modulating intersection control by sharing perception. The computational latency of complex fusion algorithms can be reduced by model compression, edge computing and hierarchical information processing to reduce real-time pressure on the algorithm, architecture and scheduling levels. Overall, the deep integration of perception and control is both an inevitable technological trend and a key path toward higher autonomy and reliability in unstructured environments.