AI & ML
Grounding Driving VLA via Inverse Kinematics
arXiv cs.RO0 views
arXiv:2605.21061v3 Announce Type: replace-cross Abstract: Counterfactual obstacle insertion exposes blind planning in representative Driving VLAs: their trajectories remain largely unchanged despite a vehicle appearing directly ahead. An empirically calibrated bias-variance analysis identifies how ego-predictable training examples admit a shortcut through ego status and commands. We address this failure with an inverse-kinematics-inspired architecture that anchors trajectory decoding on current and predicted future visual states. First, a next visual state prediction objective that requires the LLM to predict the future visual scene provides dense visual supervision and suppresses shortcut paths. Second, a separate Inverse Kinematics Network (a cross-attention-based conditional diffusion model) that takes only the current and future visual states as input is designed to suppress reliance on ego status and textual shortcuts during trajectory decoding. Our 0.5B model achieves state-of-the-art trajectory planning performance on NAVSIM-v1, NAVSIM-v2, and nuScenes using cameras alone, without LiDAR or a BEV encoder. Against the matched OpenDriveVLA* baseline, it improves NAVSIM PDMS and EPDMS by 19.0 and 20.4 points, respectively. Controlled ablations and intervention analyses support improved use of visual features, with future-state prediction sharpening hazard-specific responses.
Source:https://arxiv.org/abs/2605.21061PreviousDeformMaster: An Interactive Physics-Neural World Model for Deformable Objects from VideosNextOrbis 2: A Hierarchical World Model for Driving
Related posts
NATO Burns 300B Tokens A Month On OpenAI's Cyber Defense AI
0dev.to Machine Learning
Samsung Lets Galaxy S26 Owners Lock Apps And Track Pills In Beta
0dev.to Machine Learning
ChatGPT And Claude Can Now Shop For You, But Stores Still Block Bots
0dev.to Machine Learning
Trust Is the Deployment Obstacle: Why African AI Projects Stall After the Demo
0dev.to Machine Learning
How to Trim Audio for an Island Breeze Video — Free, in Your Browser
0dev.to Tutorial
Why Your Paycheck Never Matches Your Offer Letter
0dev.to Tutorial