Canonical Policy: Learning Canonical 3D Representation for Equivariant Policy

Visual Imitation learning has achieved remarkable progress in robotic manipulation, yet generalization to unseen objects, scene layouts, and camera viewpoints remains a key challenge. Recent advances address this by using 3D point clouds, which provide geometry-aware, appearance-invariant representations, and by incorporating equivariance into policy architectures to exploit spatial symmetries. However, existing equivariant approaches often lack interpretability and rigor due to unstructured integration of equivariant components. We introduce canonical policy, a principled framework for 3D equivariant imitation learning that unifies 3D point cloud observations under a canonical representation. We first establish a theory of 3D canonical representations, enabling equivariant observation-to-action mappings by grouping both in-distribution and out-of-distribution point clouds to a canonical representation. We then propose a flexible policy learning pipeline that leverages geometric symmetries from canonical representation and the expressiveness of modern generative models. We validate canonical policy on 12 diverse simulated tasks and 4 real-world manipulation tasks across 16 configurations, involving variations in object color, shape, camera viewpoint, and robot platform. Compared to state-of-the-art imitation learning policies, canonical policy achieves an average improvement of 18.0% in simulation and 37.6% in real-world experiments, demonstrating superior generalization capability and sample efficiency. For more details, please refer to the project website: https://zhangzhiyuanzhang.github.io/cp-website/.

ManiFeel: Benchmarking and Understanding Visuotactile Manipulation Policy Learning

Supervised visuomotor policies have shown strong performance in robotic manipulation but often struggle in tasks with limited visual input, such as operations in confined spaces, dimly lit environments, or scenarios where perceiving the object's properties and state is critical for task success. In such cases, tactile feedback becomes essential for manipulation. While the rapid progress of supervised visuomotor policies has benefited greatly from high-quality, reproducible simulation benchmarks in visual imitation, the visuotactile domain still lacks a similarly comprehensive and reliable benchmark for large-scale and rigorous evaluation. To address this, we introduce ManiFeel, a reproducible and scalable simulation benchmark for studying supervised visuotactile manipulation policies across a diverse set of tasks and scenarios. ManiFeel presents a comprehensive benchmark suite spanning a diverse set of manipulation tasks, evaluating various policies, input modalities, and tactile representation methods. Through extensive experiments, our analysis reveals key factors that influence supervised visuotactile policy learning, identifies the types of tasks where tactile sensing is most beneficial, and highlights promising directions for future research in visuotactile policy learning. ManiFeel aims to establish a reproducible benchmark for supervised visuotactile policy learning, supporting progress in visuotactile manipulation and perception. To facilitate future research and ensure reproducibility, we will release our codebase, datasets, training logs, and pretrained checkpoints. Please visit the project website for more details: https://zhengtongxu.github.io/manifeel-website/

ChicGrasp: Imitation-Learning based Customized Dual-Jaw Gripper Control for Delicate, Irregular Bio-products Manipulation

Automated poultry processing lines still rely on humans to lift slippery, easily bruised carcasses onto a shackle conveyor. Deformability, anatomical variance, and strict hygiene rules make conventional suction and scripted motions unreliable. We present ChicGrasp, an end--to--end hardware--software co-design for this task. An independently actuated dual-jaw pneumatic gripper clamps both chicken legs, while a conditional diffusion-policy controller, trained from only 50 multi--view teleoperation demonstrations (RGB + proprioception), plans 5 DoF end--effector motion, which includes jaw commands in one shot. On individually presented raw broiler carcasses, our system achieves a 40.6\% grasp--and--lift success rate and completes the pick to shackle cycle in 38 s, whereas state--of--the--art implicit behaviour cloning (IBC) and LSTM-GMM baselines fail entirely. All CAD, code, and datasets will be open-source. ChicGrasp shows that imitation learning can bridge the gap between rigid hardware and variable bio--products, offering a reproducible benchmark and a public dataset for researchers in agricultural engineering and robot learning.

DiffOG: Differentiable Policy Trajectory Optimization with Generalizability

Imitation learning-based visuomotor policies excel at manipulation tasks but often produce suboptimal action trajectories compared to model-based methods. Directly mapping camera data to actions via neural networks can result in jerky motions and difficulties in meeting critical constraints, compromising safety and robustness in real-world deployment. For tasks that require high robustness or strict adherence to constraints, ensuring trajectory quality is crucial. However, the lack of interpretability in neural networks makes it challenging to generate constraint-compliant actions in a controlled manner. This paper introduces differentiable policy trajectory optimization with generalizability (DiffOG), a learning-based trajectory optimization framework designed to enhance visuomotor policies. By leveraging the proposed differentiable formulation of trajectory optimization with transformer, DiffOG seamlessly integrates policies with a generalizable optimization layer. Visuomotor policies enhanced by DiffOG generate smoother, constraint-compliant action trajectories in a more interpretable way. DiffOG exhibits strong generalization capabilities and high flexibility. We evaluated DiffOG across 11 simulated tasks and 2 real-world tasks. The results demonstrate that DiffOG significantly enhances the trajectory quality of visuomotor policies while having minimal impact on policy performance, outperforming trajectory processing baselines such as greedy constraint clipping and penalty-based trajectory optimization. Furthermore, DiffOG achieves superior performance compared to existing constrained visuomotor policy.

VILP: Imitation Learning with Latent Video Planning

In the era of generative AI, integrating video generation models into robotics opens new possibilities for the general-purpose robot agent. This paper introduces imitation learning with latent video planning (VILP). We propose a latent video diffusion model to generate predictive robot videos that adhere to temporal consistency to a good degree. Our method is able to generate highly time-aligned videos from multiple views, which is crucial for robot policy learning. Our video generation model is highly time-efficient. For example, it can generate videos from two distinct perspectives, each consisting of six frames with a resolution of 96x160 pixels, at a rate of 5 Hz. In the experiments, we demonstrate that VILP outperforms the existing video generation robot policy across several metrics: training costs, inference speed, temporal consistency of generated videos, and the performance of the policy. We also compared our method with other imitation learning methods. Our findings indicate that VILP can rely less on extensive high-quality task-specific robot action data while still maintaining robust performance. In addition, VILP possesses robust capabilities in representing multi-modal action distributions. Our paper provides a practical example of how to effectively integrate video generation models into robot policies, potentially offering insights for related fields and directions. For more details, please refer to our open-source repository https://github.com/ZhengtongXu/VILP.

UniT: Unified Tactile Representation for Robot Learning

UniT is a novel approach to tactile representation learning, using VQVAE to learn a compact latent space and serve as the tactile representation. It uses tactile images obtained from a single simple object to train the representation with transferability and generalizability. This tactile representation can be zero-shot transferred to various downstream tasks, including perception tasks and manipulation policy learning. Our benchmarking on an in-hand 3D pose estimation task shows that UniT outperforms existing visual and tactile representation learning methods. Additionally, UniT's effectiveness in policy learning is demonstrated across three real-world tasks involving diverse manipulated objects and complex robot-object-environment interactions. Through extensive experimentation, UniT is shown to be a simple-to-train, plug-and-play, yet widely effective method for tactile representation learning. For more details, please refer to our open-source repository https://github.com/ZhengtongXu/UniT and the project website https://zhengtongxu.github.io/unifiedtactile.github.io/.

LeTac-MPC: Learning Model Predictive Control for Tactile-reactive Grasping

Grasping is a crucial task in robotics, necessitating tactile feedback and reactive grasping adjustments for robust grasping of objects under various conditions and with differing physical properties. In this paper, we introduce LeTac-MPC, a learning-based model predictive control (MPC) for tactile-reactive grasping. Our approach enables the gripper grasp objects with different physical properties on dynamic and force-interactive tasks. We utilize a vision-based tactile sensor, GelSight, which is capable of perceiving high-resolution tactile feedback that contains the information of physical properties and states of the grasped object. LeTac-MPC incorporates a differentiable MPC layer designed to model the embeddings extracted by a neural network (NN) from tactile feedback. This design facilitates convergent and robust grasping control at a frequency of 25 Hz. We propose a fully automated data collection pipeline and collect a dataset only using standardized blocks with different physical properties. However, our trained controller can generalize to daily objects with different sizes, shapes, materials, and textures. Experimental results demonstrate the effectiveness and robustness of the proposed approach. We compare LeTac-MPC with two purely model-based tactile-reactive controllers (MPC and PD) and open-loop grasping. Our results show that LeTac-MPC has the best performance on dynamic and force-interactive tasks and the best generalization ability. We release our code and dataset at https://github.com/ZhengtongXu/LeTac-MPC.

LeTO: Learning Constrained Visuomotor Policy with Differentiable Trajectory Optimization

This paper introduces LeTO, a method for learning constrained visuomotor policy via differentiable trajectory optimization. Our approach uniquely integrates a differentiable optimization layer into the neural network. By formulating the optimization layer as a trajectory optimization problem, we enable the model to end-to-end generate actions in a safe and controlled fashion without extra modules. Our method allows for the introduction of constraints information during the training process, thereby balancing the training objectives of satisfying constraints, smoothing the trajectories, and minimizing errors with demonstrations. This "gray box" method marries the optimization-based safety and interpretability with the powerful representational abilities of neural networks. We quantitatively evaluate LeTO in simulation and on the real robot. In simulation, LeTO achieves a success rate comparable to state-of-the-art imitation learning methods, but the generated trajectories are of less uncertainty, higher quality, and smoother. In real-world experiments, we deployed LeTO to handle constraints-critical tasks. The results show the effectiveness of LeTO comparing with state-of-the-art imitation learning approaches. We release our code at https://github.com/ZhengtongXu/LeTO.