This work proposes TIDAL (Temporally Interleaved Diffusion and Action Loop), a hierarchical framework that decouples semantic reasoning from high-frequency actuation in large-scale Vision-Language-Action models and introduces a temporally misaligned training strategy to handle the resulting latency shift.
Yuteng Sun, Hao-Ran Wang, Ruofei Bai et al.· arXiv.org· 5 citations
It is confirmed that, in modulation-conditioned diffusion policies, task adaptation can be routed entirely through the conditioning pathway, leaving a frozen, observation-free backbone that serves as a reusable trajectory prior, and a 5M-parameter MLP backbone matches or exceeds both the 244M U-Net and the transformer...
Jian Zhou, Si-Hao Lin, Shuai Fu et al.· 0 citations
Path planning in complex environments poses significant challenges, particularly in achieving time efficiency while maintaining a fair success rate and path cost. To address these issues, we introduce a novel path-planning algorithm, Zonal RL-RRT, that leverages kd-tree partitioning to segment the map into zones while...
Amir Tahmasbi, MohammadSaleh Faghfoorian, Aniket Bera· 0 citations
This paper studies how privileged information about a physical system affects the discovery of high-performing policies when training a reinforcement learning agent directly in the physical world, and proves that privileged observations can be decisive for policy discovery even when the resulting actions can be replaye...
Antonio Terpin, Raffaello D'Andrea· 3 citations
Reach audiences
Advertise in front of researchers, engineers, and readers.
In this paper, we describe a representation for spatial information, called the stochastic map, and associated procedures for building it, reading information from it, and revising it incrementally as new information is obtained. The map contains the estimates of relationships among objects in the map, and their uncert...
Randall Smith, Matthew Self, Peter Cheeseman· 0 citations
Slip detection is fundamental to dexterous manipulation, yet existing systems often lack precise characterization of detection latency and cross-platform generalization. We present SlipSense, a multimodal tactile slip-detection framework built on TacV5, a compact sensor integrating a $32 \times 32$ piezoresistive array...
Tong Jian, A. Kumar, Xin-Yi Li et al.· 0 citations
Bench2Dex is meant as a platform for studying visuo-tactile learning across dexterous hands and offers a shared setting for algorithm development while tactile hardware and simulation models are still evolving.
Zhen-Jie Yang, Yi-Deng Zhang, Dong-Jie Zhang et al.· 1 citation
DiffAdapterVLA is introduced, which realizes Planning in the Backbone: it injects explicit trajectory tokens into selected VLM late layers, bringing trajectory state into backbone forward computation, where it co-evolves with driving conditions at different depths.
Chang-Xin Lu, Xiao-Liang Meng, Yu Wu et al.· 0 citations
A legal planning stack that employs Defeasible Deontic Logic (DDL) to constrain a motion planner and leverages learned world models to plan and to provide legal context, enabling ex ante governance that intervenes before an illegal action is executed.
Dylan Waldner, Yiannis Kantaros, Guido Governatori et al.· 0 citations
It is shown that a frozen generative robot policy---based on either diffusion or flow matching---can be steered at inference time to respect such lexicographically ordered deployment objectives, and a controlled manipulation study shows that the dynamic barrier reaches comparable best performance over a substantially w...
Results reveal when multimodal inputs support task success and how their contributions become conditionally coupled during closed-loop execution and how their contributions become conditionally coupled during closed-loop execution.
It is argued that the composition of independent sub-policies deserves to be treated as a research problem in its own right, rather than as an implementation detail left to whatever mechanism happens to be at hand.
Daniel Gigliotti, Flavio Maiorana, Fabio Patrizi et al.· 0 citations
A new system identifies and corrects unreliable object descriptions stored in a robot’s memory, improving the robot’s ability to interact with its environment.
Robots are getting smarter, but how can their hardware match that growth? New Microsoft Research findings show that moving AI inference beyond the robot can improve task success, boost efficiency, and support more advanced physical AI workloads. The post Offloaded inference for real-world physical AI robotics appeared first on Microsoft Research.
