analiza aiqbench
AdaReP introduces a training-free wrapper that optimizes replanning tolerance in neural world-model predictive control, significantly reducing computational overhead without altering the underlying model. By dynamically adjusting to prediction mismatch and local dynamics sensitivity, the method achieves over 80% fewer queries in physical robot experiments while maintaining task performance. This advancement addresses critical efficiency bottlenecks in deploying complex AI-driven robotic controllers.
AdaReP:Adaptive Re-Planning under Model Mismatch for Neural World-Model Predictive Control
arXiv:2606.23079v2 Announce Type: replace-cross Abstract: Neural world models coupled with model predictive control (MPC) replan at every environment step to bound accumulated prediction error, but this incurs substantial computational overhead. Reusing a cached plan reduces this overhead, yet its effectiveness depends on how prediction mismatch propagates through the local dynamics. We analyze this trade-off with a perturbation-based dynamic-regret framework and show that stale-plan penalties scale with the reuse tolerance, the accumulated mismatch since the last replanning step, and the local dynamics sensitivity. Based on this structure, we propose AdaReP, a training-free wrapper that adapts the replanning tolerance online using the current deviation from the cached rollout and a local sensitivity estimate, without modifying the learned world model or planner. Across image-space planning, latent-space control, and real-world robotic manipulation, AdaReP substantially reduces planner-side computation while maintaining comparable task performance, including over 80% fewer queries on a 50-trial physical robot study.
Abstract:Neural world models coupled with model predictive control (MPC) replan at every environment step to bound accumulated prediction error, but this incurs substantial computational overhead. Reusing a cached plan reduces this overhead, yet its effectiveness depends on how prediction mismatch propagates through the local dynamics. We analyze this trade-off with a perturbation-based dynamic-regret framework and show that stale-plan penalties scale with the reuse tolerance, the accumulated mismatch since the last replanning step, and the local dynamics sensitivity. Based on this structure, we propose AdaReP, a training-free wrapper that adapts the replanning tolerance online using the current deviation from the cached rollout and a local sensitivity estimate, without modifying the learned world model or planner. Across image-space planning, latent-space control, and real-world robotic manipulation, AdaReP substantially reduces planner-side computation while maintaining comparable task performance, including over 80% fewer queries on a 50-trial physical robot study.
Accepted at ICANN 2026 as oral presentation. This arXiv version contains supplementary materials and appendices that are omitted from the conference version due to space limitations
Robotics (cs.RO); Artificial Intelligence (cs.AI)
arXiv:2606.23079 [cs.RO]
(or arXiv:2606.23079v2 [cs.RO] for this version)
https://doi.org/10.1007/978-3-032-38404-1_18 DOI(s) linking to related resources
Text integral sub licența deschisă a publicației · 189 cuvinte
Subiecte conexe
[AINews] not much happened today
Latent Space · 10 sept. 2026
The Machines that Make the Machines
NVIDIA · 7 oct. 2026
[AINews] Reality Checks on AI News (Yegge shuts down Gas Town, Databricks’ +60% Astra cost)
Latent Space · 17 sept. 2026
[AINews] Here are 6 Clones of Jev in 2 days
Latent Space · 19 sept. 2026