Yongji Fu 符永骥

For a humanoid face robot to appear approachable and believable in education, companionship, and performance settings, it needs to generate full-face dynamic expressions that engage the brows, eyes, lips, and cheeks. Existing systems typically treat the robot face as a downstream execution target of human facial animation, relying on low-dimensional expression parameters or pixel-video intermediates — which splits expression generation from physical execution: the former struggles to carry the diverse full-face motion that accompanies a single utterance, while the latter places the real robot under an out-of-distribution visual target and still requires extraction, smoothing, and mapping before it reaches the actuators. This work proposes a text- and audio-driven, high-DoF-density biomimetic face-robot system. On the hardware side, a flexible tendon-sheath transmission network decouples the servos from the facial actuation points, packing 34 servo actuators into a compact, approachable head only 19.5 cm tall. On the algorithm side, we learn a robot-aware motion latent interface and generate full-face motion directly in this space, so that text controls emotion and action semantics, audio constrains lip-sync, and robot image observations anchor the interface to the real hardware. The system avoids first generating pixel video or reconstructing the robot’s face at inference time, reducing intermediate artifacts and extra latency. Experiments show the system generates natural, controllable speaking expressions and provides a unified interface for robot expression driving that stays stable across performers.

Large-scale Vehicle Routing Problems (VRPs) face two long-standing difficulties. On the one hand, many scalable methods rely on partitioning, local candidate restriction, or staged decision making to control computation, which weakens their modeling of global structure. On the other hand, although many methods introduce search at test time to improve the final solution, search is still typically used only as a one-shot post-processing step after model prediction. The model makes a prediction, search repairs it, and little sustained feedback is formed between the two. Improved structural states are rarely fed back to the model for subsequent inference, and high-quality search solutions are seldom turned into later training supervision.

To address this issue, we propose LSL (Learning to Search, Searching to Learn), a closed-loop learning-search framework for large-scale VRPs. LSL first predicts search-friendly structural priors on a sparse candidate graph, and search then iteratively refines the current solution under the guidance of these priors. In turn, search does not leave the system after one round of refinement. At inference time, the structural states returned by search are fed back to the model for the next round of prediction, while at training time, multiple high-quality search solutions are reorganized into row-wise soft targets for model update. In this way, learning tells search where to explore, and search tells the model which structures are worth learning. Experiments show that LSL achieves strong scalability, efficiency, and solution quality across multiple large-scale VRP benchmarks.

Optical tactile sensors differ widely in geometry, illumination, and readout, so a representation or policy learned on one sensor rarely transfers to another. This forces practitioners to re-collect data and re-train for every new fingertip. TouchBridge addresses this with a reversible canonical tactile space: a shared latent that any sensor can be mapped into and, crucially, mapped back out of, so tactile signals can be translated across heterogeneous sensors rather than merely embedded.

Alignment is learned through multi-bridge alignment — instead of one monolithic encoder, each sensor connects to the canonical space through its own invertible bridge, and the bridges are trained jointly so that the canonical space stays sensor-agnostic while remaining decodable back to each native sensor domain. This lets a manipulation policy or representation trained on one sensor operate on another with little or no new tactile data.

We evaluate on cross-sensor manipulation, transferring tactile representations and contact-driven control across sensors, and show that the reversible canonical space preserves the contact information needed for downstream manipulation while enabling sensor-to-sensor translation.

Tactile sensing provides robots with direct information about physical properties through contact, yet most existing methods describe tactile data using predefined attribute labels with limited semantic flexibility. Aligning tactile signals with human language enables richer, concept-level representations.

In this work, we propose a transformer-based tactile–language framework that structures the shared embedding space as a manipulable concept space using steering vectors. These vectors encode tactile properties as semantic directions, providing explicit semantic control under limited supervision. The framework supports two complementary tasks out of the same latent space: given a free-form natural language query describing desired tactile properties, the robot retrieves the most relevant material from its tactile experience; and after contacting a surface physically, the robot generates a natural language description of its tactile properties.

Experimental results show that the framework effectively retrieves tactile representations from free-form natural language and generates meaningful tactile descriptions grounded in tactile perception, supporting more effective human–robot interaction.

Substitution boxes (S-boxes) lie at the heart of modern symmetric ciphers, and the cryptographic strength of a cipher depends heavily on the nonlinearity, differential uniformity, and unpredictability of its S-box. Static S-boxes, however, are vulnerable to algebraic and side-channel attacks once their structure is fixed. This work proposes a dynamic S-box construction scheme that combines a two-dimensional chaotic map with irreducible polynomials over the finite field .

The chaotic map provides key-dependent sensitivity and a vast parameter space, so that each key induces a distinct S-box; irreducible polynomials over supply algebraic structure that bounds worst-case cryptographic metrics. We conduct ablation experiments to evaluate the generated S-boxes on the standard battery of cryptographic criteria — nonlinearity, strict avalanche criterion (SAC), bit independence criterion (BIC), and differential / linear approximation probabilities — and on generation efficiency. The constructed S-boxes consistently meet or exceed the requirements for cryptographic use.

We further integrate the dynamic S-box into a block-cipher image encryption pipeline, and show on standard test images that it delivers strong pixel-level confusion and diffusion, near-uniform ciphertext histograms, high NPCR/UACI scores against plaintext-sensitivity attacks, and robust resistance to differential, statistical, and brute-force attacks.