Event
Shared-module recursive architecture improves generalization in universal morphology control
1 reports · 1 sources
Laurens Engwegen and other authors released a paper proposing a Transformer-based architecture with shared modular recursion for universal morphology robot control. The authors state that contextual features are often only partially observable but can be recovered through interactions between modules, thereby improving multi-robot control and generalization to contexts unseen during training. The study evaluates generalization performance on a set of MuJoCo robots. Results show that in four environments, zero-shot generalization performance is significantly improved for robots with unseen dynamics, kinematics, and topology.
Report timeline
- Shared Modular Recurrence in Contextual MDPs for Universal Morphology Control
The paper proposes a Transformer-based architecture with shared modular recurrence for universal morphology robot control and evaluates its generalization performance on a set of MuJoCo robots. The authors note that contextual features are often only partially available but can be recovered through modular interactions, improving multi-robot control and generalization to contexts unseen during training. Results show substantial improvement in zero-shot generalization performance on robots with unseen dynamics, kinematics, and topologies across four environments.
arXiv cs.RO ·
Related topics
Interest trend
- Current
- 3
- Peak
- 4Nov 4 01:00
- Past 24 hours
- –
The trend compares participants observed throughout the same period, so its range may be smaller than the current interest count. Move or tap to inspect each hour; use the arrow keys on a keyboard.