Organizations: College of Information Science and Engineering, Ritsumeikan University, Osaka, Japan · College of Information Science and Engineering, Ritsumeikan University, Osaka, Japan. · Production and Technology Department, NIPPN CORPORATION, Tokyo, Japan · Production & Technology Department, NIPPN CORPORATION, Tokyo, Japan. · University of Osaka, Osaka, Japan · University of Osaka, Osaka, Japan. · National Institute of Technology, Ichinoseki College, Iwate, Japan · National Institute of Technology, Ichinoseki College, Iwate, Japan.
Robotic systems operating in real-world environments often suffer from concept shift, where the input-output relationship changes due to latent environmental factors that are not directly observable. Conventional adaptation methods update model parameters, which may cause catastrophic forgetting and incur high computational cost. This paper proposes a latent Trend ID-based framework for few-shot adaptation in non-stationary environments. Instead of modifying model weights, a low-dimensional environmental state, referred to as the Trend ID, is estimated via backpropagation while the model parameters remain fixed. To prevent overfitting caused by per-sample latent variables, we introduce temporal regularization and a state transition model that enforces smooth evolution of the latent space. Experiments on a quantitative food grasping task demonstrate that the learned Trend IDs are distributed across distinct regions of the latent space with temporally consistent trajectories, and that few-shot adaptation to unseen environments is achieved without modifying model parameters. The proposed framework provides a scalable and interpretable solution for robotics applications operating across diverse and evolving environments.
Figures & tables
Fig. 1 : Overview of the proposed Trend ID framework. Latent environment factors—such as robot configuration, target material properties, and ambient conditions—are embedded as low-dimensional Trend IDs in a continuous latent space. The deep model receives both observable sensor data and the estimated Trend ID to output a predicted distribution adapted to the current environment.
Fig. 2 : Architecture of the proposed Trend-ID based adaptation framework. (a) During training, both the fully connected layer G and the Trend IDs {zi} are jointly optimized while the feature extractor F is held fixed. (b) During testing, both F and G are frozen, and only the Trend ID ztest is optimized from a small number of observed samples via backpropagation.
Fig. 3 : Architecture of the proposed framework. The frozen feature extractor F and trainable FC layer G receive the input and Trend ID to output a predicted distribution. Trend IDs are derived from the state transition model (left), and trainable parameters (shaded) are optimized via backpropagation (dashed arrows).
Fig. 4 : Structured trend space constructed during training. Each trajectory corresponds to a time sequence collected under a specific factory and date condition. Different colors denote different environmental conditions (factory/date/object type). The smooth trajectories indicate that the temporal constraint successfully enforces continuity in the latent space.
Fig. 5 : Visualization of the estimated Trend IDs in the test environment after few-shot adaptation. Each point represents a Trend ID in the 2D latent space ( px , py ). The estimated Trend IDs are placed within the region spanned by the training latent space, suggesting that the model integrates the unseen environment into the existing latent structure using a small number of samples.
Real-world dynamics shifts pose a critical challenge for reinforcement learning in robotics, as policies tightly coupled to nominal environments often fail catastrophically when physical conditions change. Most existing methods rely on encoding explicitly identified physical parameters into a latent context, a parameter-centric paradigm that depends on pre-specified axes of variation and becomes brittle under unmodeled or compound dynamics changes. We revisit dynamics adaptation from an outcome-centric perspective: rather than telling policies what the dynamics are, we enable them to learn how dynamics affect interaction outcomes. Theoretically, this is grounded in a monotonic relationship between target-domain regret and the Lipschitz constant of a trajectory dynamics encoder. Practically, this constant can be upper-bounded through contrastive learning, yielding a smooth, task-relevant latent topology without privileged dynamics information. On MuJoCo benchmarks, our method consistently outperforms parameter-centric baselines under severe dynamics shifts, including unmodeled and time-varying parameters, while also improving in-distribution stability and latent interpretability. Overall, these results validate that controlling latent geometry is a principled mechanism for robust adaptation.
Zhiming Xu, Weitao Zhou, Xianghui Pan +4
School of Computer Science and Engineering, Tongji University, Shanghai, China · School of Vehicle and Mobility, Tsinghua University, Beijing, China · Simple AI, Beijing, China +1
Vision-Language-Action (VLA) models provide a versatile foundation for general robotic manipulation, yet they exhibit significant brittleness when confronted with novel task distributions. While In-Context Imitation Learning (ICIL) offers a training-free alternative, existing frameworks suffer from an adaptation bottleneck that hinders the effective translation of expert context to executable actions. This failure originates from superficial retrieval mechanisms and an inherent behavioral inertia that anchors the policy to its pre-trained priors. To address these limitations, we present RA-VLA, a retrieval-augmented VLA framework that integrates behavior-aligned context retrieval with a grounded execution pipeline. By enforcing faithful adherence to functional cues within a scalable architecture, RA-VLA facilitates seamless task adaptation while preserving inference efficiency. Our empirical evaluations across the LIBERO benchmark and a real-world UR5e environment demonstrate that RA-VLA achieves superior success rates and computational efficiency, establishing a robust framework for training-free robotic adaptation.
Sanghwan Jang, Minjin Jeon, Minsoo Kim +3
Department of Computer Science and Engineering, POSTECH, Pohang, South Korea · Graduate School of Artificial Intelligence, POSTECH, Pohang, South Korea
Reinforcement-learning policies for robot locomotion can fail at deployment when operating conditions shift (e.g., rough or slippery terrain, payload changes, sensing degradation), and retraining or manual recalibration is often infeasible under on-device latency and memory constraints. We study test-time adaptation (TTA) for robotic control: adapting a pretrained policy online and in a self-supervised manner during execution, without access to rewards. Our starting point is a lightweight baseline that updates only the actor using a frozen pretrained critic as a deployment-time performance surrogate. Building on this, we introduce TEMPO, a test-time entropy-regularized, memory-guided policy optimization method that stabilizes critic-driven updates under severe shift. TEMPO combines (i) an entropy-promoting regularizer on the critic's predictive distribution to mitigate overconfident extrapolation, and (ii) a score-based memory, inspired by prioritized replay, that prioritizes low-value ("hard") observations and applies importance sampling to focus limited computation on the most informative deployment conditions. Across Go1, Go2, and G1 in MuJoCo Playground under four representative shifts and multiple architectures, TEMPO yields consistent gains, reaching up to 80% improvement in a single-device setting. We further validate on a physical Unitree Go2 with an added calf payload, where fully on-device adaptation improves performance over the non-adapted baseline across real-world episodes.
Motasem Alfarra, Pietro Mazzaglia, Markus Peschl +2