Track 03 of 09
Physical AI and Embodied Intelligence
Learning and reasoning where the agent has a body: perception–action loops, sim-to-real transfer, world models, manipulation, locomotion, and vision-language-action models.
Scope
Learning and reasoning where the agent has a body: perception–action loops, sim-to-real transfer, world models, manipulation, locomotion, and vision-language-action models.
Research topics
Topics include, but are not limited to:
- Vision-language-action (VLA) models
- Imitation and demonstration learning
- Sim-to-real transfer and domain randomisation
- World models and predictive representations
- Dexterous manipulation and grasping
- Legged and soft-body locomotion
- Tactile and multimodal sensing
- Foundation models for robotics
- Data-efficient embodied learning
- Embodied benchmarks and simulators
Emerging challenges
- The reality gap
- Scarcity of real-world interaction data
- Compute and latency limits on-board
- Generalisation to unseen objects and environments
- Safe exploration on physical hardware
- Standardised evaluation of embodied agents
Industry applications
- Bin picking and assembly
- Agricultural harvesting
- Warehouse manipulation
- Construction and demolition robotics
- Household and service robots
- Hazardous-environment inspection
Keywords
Every submission must connect to physical action, embodiment or real-world deployment, use the IEEE conference format (A4, up to 6 pages including references) and include a Deployment Risk and Limitations statement (max. 200 words). Submission guidelines →