3D human motion forecasting aims to enable autonomous applications. Estimating uncertainty for each prediction (i.e., confidence based on probability density or quantile) is essential for safety-critical contexts like human-robot collaboration to minimize risks. However, existing diverse motion fore-casting approaches struggle with uncertainty quantification due to implicit probabilistic representations hindering uncertainty modeling. We propose ProbHMI, which introduces invertible networks to parameterize poses in a disentangled latent space, enabling probabilistic dynamics modeling. A forecasting module then explicitly predicts future latent distributions, allowing effective uncertainty quantification. Evaluated on benchmarks, ProbHMI achieves strong performance for both deterministic and diverse prediction while validating uncertainty calibration, critical for risk-aware decision making.

Directly learning the swimming behaviors of real fish can significantly enhance the swimming performance of bionic robotic fish. This paper presents a novel transfer learning method based on a dynamic trajectory control approach for the robotic fish to learn swimming skills from real fish. First, we develop a fish motion capture system and a crucial motion extraction approach to realize precise decomposition of fish motions and collect abundant meaningful features from a snakehead fish as pre-training data. Next, we construct a two-stage transfer learning method based on Deep Deterministic Policy Gradient (DDPG), including an offline and an online stage. Specifically, in the offline stage, the obtained pre-training data is processed for experience learning within a DDPG-based network, whereas in the online stage, a dynamic trajectory tracking method is utilized to refine the robotic fish’s motions in real time based on the learned strategies. Experimental results on a self-developed four-joint robotic fish show that the proposed method effectively extracts and transfers biological motion features into the motion control of the robotic fish. Compared to the conventional CPG method, the proposed approach exhibits stronger acceleration capabilities and more efficient swimming, resulting in enhanced maneuverability of the robotic fish. Overall, this approach provides a technical foundation for bionic robotics to learn from nature.

To ease the difficulty of acquiring annotation labels in 3D data, a common method is using unsupervised and open-vocabulary semantic segmentation, which leverage 2D CLIP semantic knowledge. In this paper, unlike previous research that ignores the "noise" raised during feature projection from 2D to 3D, we propose a novel distillation learning framework named CUS3D. In our approach, an object-level denosing projection module is designed to screen out the "noise" and ensure more accurate 3D feature. Based on the obtained features, a multimodal distillation learning module is designed to align the 3D feature with CLIP semantic feature space with object-centered constrains to achieve advanced unsupervised semantic segmentation. We conduct comprehensive experiments in both unsupervised and open-vocabulary segmentation, and the results consistently showcase the superiority of our model in achieving advanced unsupervised segmentation results and its effectiveness in open-vocabulary segmentation.