Diffusion Policies are effective at learning closed-loop manipulation policies from human demonstrations but generalize poorly to novel arrangements of objects in 3D space, hurting real-world performance. To address this issue, we propose Spherical Diffusion Policy (SDP), an SE(3) equivariant diffusion policy that adapts trajectories according to 3D transformations of the scene. Such equivariance is achieved by embedding the states, actions, and the denoising process in spherical Fourier space. Additionally, we employ novel spherical FiLM layers to condition the action denoising process equivariantly on the scene embeddings. Lastly, we propose a spherical denoising temporal U-net that achieves spatiotemporal equivariance with computational efficiency. In the end, SDP is end-to-end SE(3) equivariant, allowing robust generalization across transformed 3D scenes. SDP demonstrates a large performance improvement over strong baselines in 20 simulation tasks and 5 physical robot tasks including single-arm and bi-manual embodiments. Code is available at https://github.com/amazon-science/Spherical_Diffusion_Policy.

Robot picking and packing tasks require dexterous manipulation skills, such as rearranging objects to establish a good grasping pose, or placing and pushing items to achieve tight packing. These tasks are challenging for robots due to the complexity and variability of the required actions. To tackle the difficulty of learning and executing long-horizon tasks, we propose a novel framework called the Multi-Head Skill Transformer (MuST). This model is designed to learn and sequentially chain together multiple motion primitives (skills), enabling robots to perform complex sequences of actions effectively. MuST introduces a “progress value” for each skill, guiding the robot on which skill to execute next and ensuring smooth transitions between skills. Additionally, our model is capable of expanding its skill set and managing various sequences of sub-tasks efficiently. Extensive experiments in both simulated and real-world environments demonstrate that MuST significantly enhances the robot's ability to perform long-horizon dexterous manipulation tasks.

Multi-fingered robotic hands can potentially offer a higher degree of dexterity for grasping a variety of parts than a traditional parallel gripper or a custom end of arm tool. However, real-time grasp planning for complex robotic hands is a challenging task. In this paper, we present real-time grasping planning algorithms that utilize two simple yet effective dimension reduction strategies. The first strategy is the “intersected volume” computation which places the robotic palm relative to the object as well as its environment. The second strategy is the “finger curling planes” which decomposes a grasp contact computation problem in high-dimensional configuration space into several independent grasp contact computation problems in low-dimensional configuration space. Our algorithms have been demonstrated with simulation, physical experiments, and robotic bin-picking and kitting applications on two robotic platforms with three different robotic hands.

Traditional industrial robots have been widely used in automotive manufacturing for nearly 30 years. However, there have been very few attempts to automate mobile robotic systems for final assembly operations, despite their potential for high flexibility and capability. This paper focuses on methods of tracking a dynamic moving vehicle that is similar to the vehicle body on a moving assembly line. We have investigated two tracking methods, one using a laser scanner and the other using a visual fiducial marker. We have also studied the tracking performance of a mobile base using the pure pursuit algorithm with low pass filtering. Experimental results are presented to illustrate the remaining main challenges in achieving robotic assembly on moving assembly lines.