Papers
8
Total Citations
76
H-Index
5
About
Donghyeon Han is a hardware architecture researcher specializing in energy-efficient processors for mobile autonomous systems, with a particular focus on simultaneous localization and mapping (SLAM), deep reinforcement learning (DRL), and 3D spatial computing. His work sits at the intersection of algorithm design and silicon implementation, targeting real-world deployment on resource-constrained platforms such as autonomous robots and augmented reality devices. Han's most influential contributions include the Space-Mate processor (18 citations), a landmark achievement combining sparse mixture-of-experts neural networks with NeRF-based SLAM in a mobile-ready chip, and OmniDRL (17–21 combined citations), a highly efficient DRL training processor featuring novel dual-mode weight compression and on-chip sparse weight transposition. His DSPU work (16 citations) further demonstrated real-time RGB-D data acquisition and 3D bounding box extraction within strict mobile power budgets. More recently, his LSPU and semantic LiDAR SLAM processors have extended these capabilities to full 360° autonomous driving perception using point neural networks and multi-level acceleration architectures. Across his publication record, Han consistently pushes the boundaries of what embedded hardware can achieve, delivering systems that balance computational throughput with extreme energy efficiency — qualities essential for the next generation of intelligent mobile platforms.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8