Researchers unveil LUCID for humanoid robots
New AI framework LUCID enables robots to handle complex tasks like humans
A new hierarchical AI framework called LUCID tackles long-horizon humanoid loco-manipulation by planning over reusable skills through imagined rollouts of a learned dynamics model. It trains a structured latent-conditioned low-level policy via adversarial imitation, then freezes it while jointly learning a high-level policy and a macro-dynamics world model that predicts temporally extended state transitions. In simulated multi-object rearrangement scenarios, LUCID improved full-task success and partial-completion rates compared to prior baseline methods.
- LUCID is a hierarchical model-based RL framework developed by researchers from the University of Manchester and Italian Institute of Technology
- It combines adversarial imitation learning with a macro-dynamics world model to plan over reusable skills via imagined rollouts
- Achieves 25-30% higher task success rates than prior methods in simulated loco-manipulation scenarios
Why It Matters
LUCID could accelerate the deployment of humanoid robots in warehouses, homes, and hazardous environments by enabling them to handle complex, multi-step tasks autonomously.