DCP-CNN-based non-cooperative spacecraft non-contact attitude estimation

A non-cooperative spacecraft refers to a space object that does not communicate with or respond to other spacecraft, ground control systems, or tracking networks. Takeover control denotes the process by which a service-spacecraft assumes full responsibility for attitude regulation and orbital adjustments of a malfunctioning or non-responsive target—providing all necessary control torques to the combined system following capture Accurate attitude estimation is a critical prerequisite for successful takeover operations, as it directly affects control performance. This study presents an enhanced attitude estimation framework for non-cooperative spacecraft by integrating multi-sensor data to exploit their complementary strengths. The Color-ICP algorithm is employed to fuse 3D point clouds with 2D thermal images for improved estimation accuracy. To address performance degradation under large rotational deviations, a neural network is introduced to compensate for the limitations of Color-ICP. Simulation results demonstrate that the proposed method effectively mitigates large-angle estimation errors. Simulation results demonstrate that it addresses the performance loss of traditional Color-ICP algorithm at large turning angles through iterative optimization of Color-ICP and neural networks, achieving accuracy and larger angular range. The proposed DCP-CNN based attitude estimation framework can provide attitude information of non-cooperative spacecraft for takeover control.