Search papers, labs, and topics across Lattice.
5
0
5
6
Smaller, fine-tuned models can outperform larger counterparts by 30% when trained to focus on planning-critical occlusions in autonomous driving scenarios.
HOLO-MPPI achieves robust motion planning across diverse scenarios without the need for per-scenario retuning, outperforming traditional methods in both efficiency and adaptability.
Motorcycles can navigate complex road hazards more safely thanks to a new system that uses VLMs to understand and map risks in a way that's tailored to their unique dynamics.
Unlock 80% faster autonomous parking by learning a single intermediate pose that simplifies complex maneuvers.
Ditching short-term motion prediction for explicit intention prediction from motion history unlocks more accurate and socially acceptable autonomous valet parking.
