MemoryVLA
What it is
MemoryVLA is a Cognition-Memory-Action framework for robotic manipulation that implements a hippocampal-like perceptual-cognitive memory. It captures temporal dependencies to enable long-horizon, temporally aware action generation in Vision-Language-Action (VLA) models.
The problem it solves
Standard VLA models often struggle with long-horizon tasks because they lack a mechanism to maintain and utilize temporal memory for complex, multi-step robotic actions.
What you could use it for
- Long-horizon robotic manipulation tasks
- Temporally aware action generation
- Robotic manipulation in complex environments (e.g., Libero, SimplerEnv)
- Simultaneous multi-task robotic learning
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.