UfM* employs Gaussian mixture models for single-inference multiview disagreement-based depth uncertainty estimation, achieving high accuracy with minimal energy.

Soumya Sudhakar, Sertac Karaman, Vivienne Sze

AwareVLN introduces self-aware reasoning for VLN, achieving NE 4.02 on R2R-CE Val-Unseen, outperforming prior SOTA.

Wenxuan Guo, Xiuwei Xu, Yichen Liu et al.

GesVLA integrates gesture into Vision-Language-Action models, achieving 94.3% target grounding accuracy in complex real-world tasks.

Wenxuan Guo, Ziyuan Li, Meng Zhang et al.

League-based multi-agent RL achieves 22 m/s quadrotor racing with 50% collision reduction vs. single-agent baselines.

Ismail Geles, Leonard Bauersfeld, Markus Wulfmeier et al.

Proposed Branch-Stochastic MPC with scenario clustering improves safety and real-time performance in multi-modal uncertainty motion planning.

Zekun Xing, Ramkrishna Chaudhari, Marion Leibold et al.

Steins;Gate Drive employs structured future forecasting with latency decoupling, significantly improving autonomous driving safety and responsiveness.

Anjie Qiu, Hans D. Schotten

SOMA employs multi-view spatial memory to enable out-of-vision manipulation, boosting success rates by 10% in real-world tasks.

Pengteng Li, Weiyu Guo, He Zhang et al.

EvoScene-VLA updates scene states via actions, boosting RoboTwin task success to 88.8%.

Chushan Zhang, Ruihan Lu, Jinguang Tong et al.

Proposed a structural latent point learning method, achieving a 56% task success rate on RLBench.

Yicheng Jiang, Jiaxu Wang, Junhao He et al.

SUGAR leverages human videos to learn humanoid loco-manipulation skills without task rewards or reference motions, enabling zero-shot transfer.

Tianshu Wu, Xiangqi Kong, Yue Chen et al.

Proposes STRELGen, integrating diffusion models with spatio-temporal logic for targeted safety-critical scenario synthesis.

Lorenzo Bonin, Francesco Giacomarra, Luca Bortolussi et al.

Introduces 4DLidarOpen, a large-scale multi-modal dataset with 4D FMCW lidar velocity data, enhancing dynamic scene understanding.

Kane Qian, Xin Zhao, Yining Shi et al.

This survey categorizes four main methods—latent actions, world models, 2D cues, and 3D reconstructions—for transforming human videos into robot control knowledge, enabling scalable vision-language-action learning.

Zhiyuan Feng, Qixiu Li, Huizhi Liang et al.

DeMiAn method enhances robot policy learning with dense language annotations, boosting RoboCasa success by 5 points.

Bosung Kim, Ruiyi Wang, David Acuna et al.

NavRL++ introduces a system-level framework with perturbation-aware fine-tuning and Transformer-based temporal reasoning, achieving zero-shot sim-to-real transfer in robot navigation.

Zhefan Xu, Hanyu Jin, Kenji Shimada

CUBic employs shared codebooks for unified bimanual perception and control, achieving 12% higher success in RoboTwin benchmarks.

Xingyu Wang, Pengxiang Ding, Jingkai Xu et al.

Proposes a five-layer MLOps architecture enabling collective learning and safety assurance for autonomous vehicle fleets.

Bastian Lampe, Lutz Eckstein

SafeManip uses LTLf to evaluate temporal safety in robotic manipulation, revealing task success does not equal safe execution.

Chengyue Huang, Khang Vo Huynh, Sebastian Elbaum et al.

Real-time whole-body teleoperation using Virdyn IMU motion capture, validated on Unitree G1 robot with Sim2Sim and Sim2Real.

Hamza Ahmed Durrani, Suleman Khan

MoLA uses pretrained multimodal inverse dynamics to convert imagined future videos into executable robot actions, boosting success and generalization.

Yajie Li, Bozhou Zhang, Chun Gu et al.