Dynamic Distributed Fusion Trajectory Tracking for Complex Maneuvering Targets

This paper addresses the state estimation problem in resource-constrained distributed multi-sensor systems, focusing on the trade-off between estimation accuracy and system overhead under limited communication bandwidth and computational capabilities. Traditional multi-sensor fusion architectures often rely on predetermined static algorithms and fixed network topologies, which lack adaptability to dynamic environmental changes and task requirements, making it diffi⁃ cult to maintain an optimal balance between performance and efficiency over prolonged operation. To overcome these limi⁃ tations, this study proposes an event-triggered adaptive fusion scheduling framework. By employing an intelligent event-driven mechanism, the framework dynamically organizes sensor resources, thereby ensuring estimation accuracy while sig⁃ nificantly reducing unnecessary communication and computational costs. The event-oriented dynamic fusion unit enables the system to intelligently cluster and reconfigure sensor nodes in real time based on detected events. These nodes are tem⁃ porarily grouped into task-specific fusion sub-units tailored to the current event. This mechanism allows the system to focus on critical information, avoiding the substantial overhead associated with continuous global fusion across all nodes. For event determination, this paper investigates and compares two design strategies: one based on local innovation, where each node generates triggering events solely from its own measurements and innovation sequence; and the other based on global feedback, where the fused estimate is fed back from the fusion center to each node for decision-making. The former is fully distributed and offers stronger privacy, while the latter provides a more comprehensive system perspective, facilitating bet⁃ ter clustering decisions. Furthermore, after the temporary fusion unit is formed and performs local fusion, a hybrid fusion mechanism is introduced to conduct secondary fusion of the local results. This mechanism achieves an additional balance between estimation accuracy and fusion cost at the final output stage. A series of simulation experiments demonstrate that both event design schemes—based on local innovation and global feedback: successfully transform the fusion algorithm se⁃ lection from a static configuration into a dynamic real-time scheduling process. The system can autonomously adjust sensor clustering according to real-time task demands and resource availability, thereby sustaining an optimal trade-off between ac⁃ curacy and efficiency in dynamically changing environments. Comparative experiments further reveal that, under certain scheduling parameters, the global feedback-based design outperforms the fully distributed local innovation approach in esti⁃ mation performance due to its access to global system state information, which enables superior scheduling decisions. This advantage, however, comes at the cost of introducing periodic global communication. The proposed event-triggered adap⁃ tive fusion scheduling framework offers a flexible and efficient solution for resource-constrained distributed sensing sys⁃ tems. By dynamically scheduling sensors via event-driven mechanisms, it effectively resolves the balance between estima⁃ tion accuracy and hardware overhead in challenging applications such as trajectory tracking of highly maneuverable targets.