The increasing presence of non-cooperative targets poses significant challenges to the space environment and threatens the sustainability of aerospace operations. Accurate on-orbit perception of such targets, particularly those without cooperative markers, requires advanced algorithms and efficient system architectures. This study presents a hardware–software co-design framework for the pose estimation of non-cooperative targets. Firstly, a two-stage architecture is proposed, comprising object detection and pose estimation. YOLOv5s is modified with a Focus module to enhance feature extraction, and URSONet adopts global average pooling to reduce the computational burden. Optimization techniques, including batch normalization fusion, ReLU integration, and linear quantization, are applied to improve inference efficiency. Secondly, a customized FPGA-based accelerator is developed with an instruction scheduler, memory slicing mechanism, and computation array. Instruction-level control supports model generalization, while a weight concatenation strategy improves resource utilization during convolution. Finally, a heterogeneous DSP–FPGA system is implemented, where the DSP manages data pre-processing and result integration, and the FPGA performs core inference. The system is deployed on a Xilinx X7K325T FPGA operating at 200 MHz. Experimental results show that the optimized model achieves a peak throughput of 399.16 GOP/s with less than 1% accuracy loss. The proposed design reaches 0.461 and 0.447 GOP/s/DSP48E1 for two model variants, achieving a 2× to 3× improvement over comparable designs.
1 November 2026 in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
Neural Radiance Fields (NeRF) have emerged as a promising solution for 3D spatial modeling, scene reconstruction, and AR/VR applications. However, NeRF rendering involves frequent and irregular off-chip memory accesses, particularly during ray sampling and hash encoding stages. Although existing edge platforms offer sufficient computational capability and memory bandwidth, such irregular memory access patterns lead to inefficient hardware utilization and significant latency overheads, making real-time rendering challenging. To address these issues, we present NeMO, an end-to-end accelerator specifically designed to enhance memory access efficiency in instant 3D rendering. NeMO incorporates three key innovations: (1) a novel occupancy grid compaction technique that completely eliminates off-chip memory accesses during the ray marching phase, (2) a hierarchical subgrid hashing strategy with block-by-block ray processing to optimize hash table access patterns, combined with a two-level cache architecture that reduces off-chip memory access by over 99% during multi-resolution hash encoding, (3) an end-to-end processor tailored for the NeRF pipeline, featuring fine-grained pipelining for ray marching and backpressure-enabled pipelining for multi-resolution hash encoding. Implemented in a commercial 28nm process technology, NeMO achieves real-time rendering at 79.23 FPS on synthetic datasets. It outperforms the Nvidia Jetson AGX Orin and high-end RTX 4080 GPU by 51.78× and 6.43× in speed and 1,767.08× and 493.17× in energy efficiency, respectively. Additionally, NeMO surpasses the latest NeRF accelerator, delivering 2.64× higher speed and a 28.7× off-chip bandwidth reduction, with per-frame memory access under 30 MB.