Papers
3
Total Citations
79
H-Index
2
About
Jake Olkin is a robotics researcher whose work spans deformable object manipulation, legged locomotion, and multi-agent information gathering. He is best known for developing **SoftGym** (68 citations), a benchmark that standardized evaluation of deep reinforcement learning for manipulating soft, deformable objects—a notoriously difficult problem due to high-dimensional state spaces and complex dynamics. This contribution has become a foundational tool for researchers tackling real-world tasks like cloth folding or rope manipulation. In legged robotics, Olkin introduced a **Workspace CPG with Body Pose Control** (9 citations) that enables stable, directed vision during omnidirectional locomotion on high-degree-of-freedom robots, solving the challenge of gaze control for active perception without dedicated pan-tilt units. Most recently, his work on **Multi-Agent Vulcan** (2024) advances adaptive sampling by proposing an information-driven multi-agent pathfinding approach for autonomous vehicles exploring unknown environments. Olkin’s research consistently bridges simulation and real-world deployment, providing both benchmarks and algorithms that enable robots to interact with complex, unstructured environments. His work is particularly relevant for students and researchers interested in the intersection of reinforcement learning, perception, and multi-robot coordination.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3