Jiankun Wang

Uncovering a coherent long-tail structure in urban navigation reveals critical safety scenarios that traditional data approaches often miss.

Achieving a 2dB PSNR improvement, EmbodiedVAE transforms how robots learn and execute manipulation tasks by providing compact and controllable latent representations.

Achieving high accuracy in luggage trolley pose estimation while slashing computational costs and latency is now possible with a unified approach.

Dysco cuts training loss by up to 9 times and boosts federated learning performance by dynamically aligning client-specific subspaces, tackling a critical source of instability in LoRA aggregation.

Realistic pedestrian simulations can significantly improve the training and evaluation of human-aware navigation systems, bridging the gap between simulated and real-world interactions.

Robots can now navigate crowded spaces more effectively by understanding human intentions, thanks to a new method that integrates rich visual cues into their decision-making process.

Robots can now autonomously figure out where to make contact for complex manipulations, opening the door to more versatile and robust automation.

Key contribution not extracted.