UAV Formation Reconstruction Using Cross-Entropy Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current formation reconstruction algorithms for unmanned device clusters, particularly in leader-follower modes, struggle with interference conditions, leading to instability and collision issues as the cluster size increases, exemplified by errors in large-scale UAV formations.
Innovation Solution
A multi-body formation reconstruction method that reads a preset target formation, establishes probability density functions for both target and current formations, and adjusts object positions to minimize cross entropy, allowing real-time formation adjustment without pre-defined leader-follower relationships and path planning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If leader-follower mode is used for formation control, then formation structure can be maintained, but the system becomes difficult to deal with interference conditions and loses stability
Solution Approach 1:
The patent extracts and eliminates the leader-follower hierarchical structure from the formation control system. By removing the designated leader role and the corresponding path planning dependencies, the system avoids the vulnerability where leader failure causes cascade confusion. Each unmanned device independently computes its position based on probability density functions, ensuring the system maintains stability even when individual devices fail or experience interference.
Solution Approach 2:
The patent transforms the control approach from deterministic leader-follower positioning to probabilistic position distribution. By using probability density functions to describe formation positions and applying cross-entropy minimization, the system adapts to interference conditions dynamically. This parameter change from fixed hierarchical roles to flexible probabilistic positioning enables the formation to maintain reliability under interference.
2Reliability
If real-time path re-planning is implemented to handle formation disturbances, then formation can be restored, but the computational complexity becomes unmanageable for large clusters
Solution Approach 1:
The patent segments the complex global path re-planning problem into independent local position adjustments. Instead of computing new paths for entire formations or large clusters, each unmanned device independently adjusts its position by minimizing cross-entropy with the target probability density function. This segmentation reduces computational complexity from O(N) global planning to O(1) local adjustments for each device, making real-time restoration feasible for large clusters.
Solution Approach 2:
Each unmanned device performs self-service by independently computing its own position adjustment based on the target formation's probability density function. Devices autonomously minimize their cross-entropy with the target distribution without requiring centralized path re-planning or communication with leaders. This self-service mechanism enables scalable formation restoration for large clusters with manageable computational complexity.
3Ease of operation
If one-to-one correspondence is established between formation positions and controlled objects, then path planning can be performed, but the system requires pre-defined relationships that reduce flexibility under interference
Solution Approach 1:
The patent introduces dynamics by replacing static one-to-one correspondence with dynamic probability density functions. Formation positions are described by probabilistic distributions rather than fixed assignments, allowing the system to adapt dynamically to interference conditions. The cross-entropy minimization process dynamically adjusts device positions to match the target distribution without requiring pre-defined rigid relationships, maintaining both operational capability and flexibility.
Data Source
AI summary
The present invention relates to a multi-body formation reconstruction method for an unmanned device cluster control. The present invention comprises: step 1: reading a preset target formation; step 2: establishing a probability density function associated with the target formation; step 3: obtaining the current formation of all the controlled objects in the unmanned device cluster, and establishing a probability density function associated with the current formation; step 4: decreasing a cross entropy by adjusting the position of the controlled objects in the unmanned device cluster, and repeating executing the step 3 and the step 4 until the formation composed by the controlled objects in the unmanned device cluster is consistent with the target formation.