Papers
5
Total Citations
22
H-Index
3
About
Aleksandr Semochkin is a roboticist focused on bridging the gap between perception and manipulation in unstructured environments. His research centers on keypoint-based pose estimation, deep reinforcement learning, and continuous learning for autonomous systems. Semochkin’s most impactful work, “Unreal Mask” (8 citations), introduces a one-shot, multi-object pose estimation method using synthetic data, enabling robots to manipulate unknown objects without real-world training. He further advanced practical robotics with “Coinbot” (6 citations), applying deep reinforcement learning to safely automate the physically demanding task of moving heavy currency bags in bank cash centers—a direct response to real-world industrial needs. His earlier work on user-defined grasping via key-points (3 citations) and a continuous learning framework for object detection (3 citations) demonstrate his commitment to adaptive, memory-efficient robotic systems. Semochkin also contributed to path planning with a modified bi-directional RRT* algorithm for humanoid robots, prioritizing speed and memory efficiency in high-dimensional C-spaces. Through these contributions, Semochkin is shaping safer, more capable robots for collaborative human environments.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4
- 5