This paper addresses the problem of real-time onboard routing for dynamic low earth orbit (LEO) satellite networks. It is difficult to apply general routing algorithms to dynamic LEO networks due to the frequent changes in satellite topology caused by the disconnection between moving satellites. Deep reinforcement learning (DRL) models trained by various dynamic networks can be considered. However, since the inference process with the DRL model requires too long a computation time due to multiple convolutional layer operations, it is not practical to apply to a real-time on-board computer (OBC) with limited computing resources. To solve the problem, this paper proposes a practical co-design method with heterogeneous processors to parallelize and accelerate a part of the multiple convolutional layer operations on a field-programmable gate array (FPGA). The proposed method was tested with a real heterogeneous processor-based OBC and showed that the proposed method was about 3.10 times faster than the conventional method while achieving the same routing results.

Low-Earth-Orbit (LEO) satellite networks offer a promising avenue for achieving global connectivity, despite certain technical and economic challenges such as high implementation costs and the complexity of network management. Nonetheless, real-time routing remains challenging because of rapid topology changes and strict energy constraints. This paper proposes a GPU-accelerated Eclipse-Aware Routing (EAR) method that simultaneously minimizes hop count and balances energy consumption for real-time routing on an onboard computer (OBC). The approach first employs a Breadth-First Search (BFS)–based K-Shortest Paths (KSP) algorithm to generate candidate routes and then evaluates battery usage to select the most efficient path. In large-scale networks, the computational load of the KSP search increases substantially. Therefore, CUDA-based parallel processing was integrated to enhance performance, resulting in a speedup of approximately 3.081 times over the conventional CPU-based method. The practical applicability of the proposed method is further validated by successfully updating routing tables in a SpaceWire network.

본 논문는 다중 로봇 시스템에서 각 로봇이 독립적으로 수집한 환경 정보를 효율적으로 융합하는 방법을 제안한다. 로봇의 작은 크기로 인한 제한된 전력으로 인해 고성능 센서를 사용할 수 없으며, 저성능 센서를 사용할 수 밖에 없다. 이러한 센서들로부터 얻은 데이터를 융합하여 다중 객체의 위치를 수집한다. 수집된 정보는 메쉬 네트워크를 통해 로봇 간 주고 받는다. 그리고 허프 변환과 객체 중심 스펙트럼 매칭 방법을 활용해 두 환경 정보 간의 최적 회전을 추정한다. 데이터 양이 동일하거나 차이나는 시나리오에서 이 두 방법을 비교한 결과, 객체 중심 스펙트럼 매칭 방법이 빠른 수행시간과 높은 정확도를 보여준다.