Robert Platt

David M. Klee, Ondrej Biza, Robert Platt et al.

Predicting the pose of objects from a single image is an important but difficult computer vision problem. Methods that predict a single point estimate do not predict the pose of objects with symmetries well and cannot represent uncertainty. Alternatively, some works predict a distribution over orientations in $\mathrm{SO}(3)$. However, training such models can be computation- and sample-inefficient. Instead, we propose a novel mapping of features from the image domain to the 3D rotation manifold. Our method then leverages $\mathrm{SO}(3)$ equivariant layers, which are more sample efficient, and outputs a distribution over rotations that can be sampled at arbitrary resolution. We demonstrate the effectiveness of our method at object orientation prediction, and achieve state-of-the-art performance on the popular PASCAL3D+ dataset. Moreover, we show that our method can model complex object symmetries, without any modifications to the parameters or loss function. Code is available at https://dmklee.github.io/image2sphere.

Kevin Esslinger, Robert Platt, Christopher Amato

Real-world reinforcement learning tasks often involve some form of partial observability where the observations only give a partial or noisy view of the true state of the world. Such tasks typically require some form of memory, where the agent has access to multiple past observations, in order to perform well. One popular way to incorporate memory is by using a recurrent neural network to access the agent's history. However, recurrent neural networks in reinforcement learning are often fragile and difficult to train, susceptible to catastrophic forgetting and sometimes fail completely as a result. In this work, we propose Deep Transformer Q-Networks (DTQN), a novel architecture utilizing transformers and self-attention to encode an agent's history. DTQN is designed modularly, and we compare results against several modifications to our base model. Our experiments demonstrate the transformer can solve partially observable tasks faster and more stably than previous recurrent approaches.

Haojie Huang, Dian Wang, Xupeng Zhu et al.

Given point cloud input, the problem of 6-DoF grasp pose detection is to identify a set of hand poses in SE(3) from which an object can be successfully grasped. This important problem has many practical applications. Here we propose a novel method and neural network model that enables better grasp success rates relative to what is available in the literature. The method takes standard point cloud data as input and works well with single-view point clouds observed from arbitrary viewing directions.

Hai Nguyen, Andrea Baisero, Dian Wang et al.

Reinforcement learning in partially observable domains is challenging due to the lack of observable state information. Thankfully, learning offline in a simulator with such state information is often possible. In particular, we propose a method for partially observable reinforcement learning that uses a fully observable policy (which we call a state expert) during offline training to improve online performance. Based on Soft Actor-Critic (SAC), our agent balances performing actions similar to the state expert and getting high returns under partial observability. Our approach can leverage the fully-observable policy for exploration and parts of the domain that are fully observable while still being able to learn under partial observability. On six robotics domains, our method outperforms pure imitation, pure reinforcement learning, the sequential or parallel combination of both types, and a recent state-of-the-art method in the same setting. A successful policy transfer to a physical robot in a manipulation task from pixels shows our approach's practicality in learning interesting policies under partial observability.

Dian Wang, Jung Yeon Park, Neel Sortur et al.

Extensive work has demonstrated that equivariant neural networks can significantly improve sample efficiency and generalization by enforcing an inductive bias in the network architecture. These applications typically assume that the domain symmetry is fully described by explicit transformations of the model inputs and outputs. However, many real-life applications contain only latent or partial symmetries which cannot be easily described by simple transformations of the input. In these cases, it is necessary to learn symmetry in the environment instead of imposing it mathematically on the network architecture. We discover, surprisingly, that imposing equivariance constraints that do not exactly match the domain symmetry is very helpful in learning the true symmetry in the environment. We differentiate between extrinsic and incorrect symmetry constraints and show that while imposing incorrect symmetry can impede the model's performance, imposing extrinsic symmetry can actually improve performance. We demonstrate that an equivariant model can significantly outperform non-equivariant methods on domains with latent symmetries both in supervised learning and in reinforcement learning for robotic manipulation and control problems.

Haojie Huang, Dian Wang, Arsh Tangri et al.

Robotic pick and place tasks are symmetric under translations and rotations of both the object to be picked and the desired place pose. For example, if the pick object is rotated or translated, then the optimal pick action should also rotate or translate. The same is true for the place pose; if the desired place pose changes, then the place action should also transform accordingly. A recently proposed pick and place framework known as Transporter Net captures some of these symmetries, but not all. This paper analytically studies the symmetries present in planar robotic pick and place and proposes a method of incorporating equivariant neural models into Transporter Net in a way that captures all symmetries. The new model, which we call Equivariant Transporter Net, is equivariant to both pick and place symmetries and can immediately generalize pick and place knowledge to different pick and place poses. We evaluate the new model empirically and show that it is much more sample efficient than the non-symmetric version, resulting in a system that can imitate demonstrated pick and place behavior using very few human demonstrations on a variety of imitation learning tasks.

Ondrej Biza, Skye Thompson, Kishore Reddy Pagidi et al.

Imitation learning of robot policies from few demonstrations is crucial in open-ended applications. We propose a new method, Interaction Warping, for learning SE(3) robotic manipulation policies from a single demonstration. We infer the 3D mesh of each object in the environment using shape warping, a technique for aligning point clouds across object instances. Then, we represent manipulation actions as keypoints on objects, which can be warped with the shape of the object. We show successful one-shot imitation learning on three simulated and real-world object re-arrangement tasks. We also demonstrate the ability of our method to predict object meshes and robot grasps in the wild.

Xupeng Zhu, Dian Wang, Guanang Su et al.

Real-world grasp detection is challenging due to the stochasticity in grasp dynamics and the noise in hardware. Ideally, the system would adapt to the real world by training directly on physical systems. However, this is generally difficult due to the large amount of training data required by most grasp learning models. In this paper, we note that the planar grasp function is $\SE(2)$-equivariant and demonstrate that this structure can be used to constrain the neural network used during learning. This creates an inductive bias that can significantly improve the sample efficiency of grasp learning and enable end-to-end training from scratch on a physical robot with as few as $600$ grasp attempts. We call this method Symmetric Grasp learning (SymGrasp) and show that it can learn to grasp ``from scratch'' in less that 1.5 hours of physical robot time.

Dian Wang, Xupeng Zhu, Jung Yeon Park et al.

Although equivariant machine learning has proven effective at many tasks, success depends heavily on the assumption that the ground truth function is symmetric over the entire domain matching the symmetry in an equivariant neural network. A missing piece in the equivariant learning literature is the analysis of equivariant networks when symmetry exists only partially in the domain. In this work, we present a general theory for such a situation. We propose pointwise definitions of correct, incorrect, and extrinsic equivariance, which allow us to quantify continuously the degree of each type of equivariance a function displays. We then study the impact of various degrees of incorrect or extrinsic symmetry on model error. We prove error lower bounds for invariant or equivariant networks in classification or regression settings with partially incorrect symmetry. We also analyze the potentially harmful effects of extrinsic equivariance. Experiments validate these results in three different environments.

Ondrej Biza, Robert Platt, Jan-Willem van de Meent et al.

We study the problem of binding actions to objects in object-factored world models using action-attention mechanisms. We propose two attention mechanisms for binding actions to objects, soft attention and hard attention, which we evaluate in the context of structured world models for five environments. Our experiments show that hard attention helps contrastively-trained structured world models to learn to separate individual objects in an object-based grid-world environment. Further, we show that soft attention increases performance of factored world models trained on a robotic manipulation task. The learned action attention weights can be used to interpret the factored world model as the attention focuses on the manipulated object in the environment.

Yaoyao Qian, Xupeng Zhu, Ondrej Biza et al.

Robotic grasping in cluttered environments remains a significant challenge due to occlusions and complex object arrangements. We have developed ThinkGrasp, a plug-and-play vision-language grasping system that makes use of GPT-4o's advanced contextual reasoning for heavy clutter environment grasping strategies. ThinkGrasp can effectively identify and generate grasp poses for target objects, even when they are heavily obstructed or nearly invisible, by using goal-oriented language to guide the removal of obstructing objects. This approach progressively uncovers the target object and ultimately grasps it with a few steps and a high success rate. In both simulated and real experiments, ThinkGrasp achieved a high success rate and significantly outperformed state-of-the-art methods in heavily cluttered environments or with diverse unseen objects, demonstrating strong generalization capabilities.

David Klee, Ondrej Biza, Robert Platt et al.

Reasoning about 3D objects based on 2D images is challenging due to variations in appearance caused by viewing the object from different orientations. Tasks such as object classification are invariant to 3D rotations and other such as pose estimation are equivariant. However, imposing equivariance as a model constraint is typically not possible with 2D image input because we do not have an a priori model of how the image changes under out-of-plane object rotations. The only $\mathrm{SO}(3)$-equivariant models that currently exist require point cloud or voxel input rather than 2D images. In this paper, we propose a novel architecture based on icosahedral group convolutions that reasons in $\mathrm{SO(3)}$ by learning a projection of the input image onto an icosahedron. The resulting model is approximately equivariant to rotation in $\mathrm{SO}(3)$. We apply this model to object pose estimation and shape classification tasks and find that it outperforms reasonable baselines. Project website: \url{https://dmklee.github.io/image2icosahedral}

Colin Kohler, Robert Platt

Model-free policy learning has been shown to be capable of learning manipulation policies which can solve long-time horizon tasks using single-step manipulation primitives. However, training these policies is a time-consuming process requiring large amounts of data. We propose the Local Dynamics Model (LDM) which efficiently learns the state-transition function for these manipulation primitives. By combining the LDM with model-free policy learning, we can learn policies which can solve complex manipulation tasks using one-step lookahead planning. We show that the LDM is both more sample-efficient and outperforms other model architectures. When combined with planning, we can outperform other model-based and model-free policies on several challenging manipulation tasks in simulation.

Hai Nguyen, Andrea Baisero, David Klee et al.

Incorporating inductive biases is a promising approach for tackling challenging robot learning domains with sample-efficient solutions. This paper identifies partially observable domains where symmetries can be a useful inductive bias for efficient learning. Specifically, by encoding the equivariance regarding specific group symmetries into the neural networks, our actor-critic reinforcement learning agents can reuse solutions in the past for related scenarios. Consequently, our equivariant agents outperform non-equivariant approaches significantly in terms of sample efficiency and final performance, demonstrated through experiments on a range of robotic tasks in simulation and real hardware.

Haojie Huang, Haotian Liu, Dian Wang et al.

Many manipulation tasks require the robot to rearrange objects relative to one another. Such tasks can be described as a sequence of relative poses between parts of a set of rigid bodies. In this work, we propose MATCH POLICY, a simple but novel pipeline for solving high-precision pick and place tasks. Instead of predicting actions directly, our method registers the pick and place targets to the stored demonstrations. This transfers action inference into a point cloud registration task and enables us to realize nontrivial manipulation policies without any training. MATCH POLICY is designed to solve high-precision tasks with a key-frame setting. By leveraging the geometric interaction and the symmetries of the task, it achieves extremely high sample efficiency and generalizability to unseen configurations. We demonstrate its state-of-the-art performance across various tasks on RLBench benchmark compared with several strong baselines and test it on a real robot with six tasks.

Dian Wang, Stephen Hart, David Surovik et al.

Recent work has shown diffusion models are an effective approach to learning the multimodal distributions arising from demonstration data in behavior cloning. However, a drawback of this approach is the need to learn a denoising function, which is significantly more complex than learning an explicit policy. In this work, we propose Equivariant Diffusion Policy, a novel diffusion policy learning method that leverages domain symmetries to obtain better sample efficiency and generalization in the denoising function. We theoretically analyze the $\mathrm{SO}(2)$ symmetry of full 6-DoF control and characterize when a diffusion model is $\mathrm{SO}(2)$-equivariant. We furthermore evaluate the method empirically on a set of 12 simulation tasks in MimicGen, and show that it obtains a success rate that is, on average, 21.9% higher than the baseline Diffusion Policy. We also evaluate the method on a real-world system to show that effective policies can be learned with relatively few training samples, whereas the baseline Diffusion Policy cannot.

Ondrej Biza, Dian Wang, Robert Platt et al.

In robotics, it is often not possible to learn useful policies using pure model-free reinforcement learning without significant reward shaping or curriculum learning. As a consequence, many researchers rely on expert demonstrations to guide learning. However, acquiring expert demonstrations can be expensive. This paper proposes an alternative approach where the solutions of previously solved tasks are used to produce an action prior that can facilitate exploration in future tasks. The action prior is a probability distribution over actions that summarizes the set of policies found solving previous tasks. Our results indicate that this approach can be used to solve robotic manipulation problems that would otherwise be infeasible without expert demonstrations. Source code is available at \url{https://github.com/ondrejba/action_priors}.

Ondrej Biza, Robert Platt, Jan-Willem van de Meent et al.

Abstraction is crucial for effective sequential decision making in domains with large state spaces. In this work, we propose an information bottleneck method for learning approximate bisimulations, a type of state abstraction. We use a deep neural encoder to map states onto continuous embeddings. We map these embeddings onto a discrete representation using an action-conditioned hidden Markov model, which is trained end-to-end with the neural network. Our method is suited for environments with high-dimensional states and learns from a stream of experience collected by an agent acting in a Markov decision process. Through this learned discrete abstract model, we can efficiently plan for unseen goals in a multi-goal Reinforcement Learning setting. We test our method in simplified robotic manipulation domains with image states. We also compare it against previous model-based approaches to finding bisimulations in discrete grid-world-like environments. Source code is available at https://github.com/ondrejba/discrete_abstractions.

Ondrej Biza, Robert Platt

Abstraction of Markov Decision Processes is a useful tool for solving complex problems, as it can ignore unimportant aspects of an environment, simplifying the process of learning an optimal policy. In this paper, we propose a new algorithm for finding abstract MDPs in environments with continuous state spaces. It is based on MDP homomorphisms, a structure-preserving mapping between MDPs. We demonstrate our algorithm's ability to learn abstractions from collected experience and show how to reuse the abstractions to guide exploration in new tasks the agent encounters. Our novel task transfer method outperforms baselines based on a deep Q-network in the majority of our experiments. The source code is at https://github.com/ondrejba/aamas_19.

Haojie Huang, Owen Howell, Dian Wang et al.

Many complex robotic manipulation tasks can be decomposed as a sequence of pick and place actions. Training a robotic agent to learn this sequence over many different starting conditions typically requires many iterations or demonstrations, especially in 3D environments. In this work, we propose Fourier Transporter (FourTran) which leverages the two-fold SE(d)xSE(d) symmetry in the pick-place problem to achieve much higher sample efficiency. FourTran is an open-loop behavior cloning method trained using expert demonstrations to predict pick-place actions on new environments. FourTran is constrained to incorporate symmetries of the pick and place actions independently. Our method utilizes a fiber space Fourier transformation that allows for memory-efficient construction. We test our proposed network on the RLbench benchmark and achieve state-of-the-art results across various tasks.

Xupeng Zhu, David Klee, Dian Wang et al.

Recent advances in Keyframe Imitation Learning (IL) have enabled learning-based agents to solve a diverse range of manipulation tasks. However, most approaches ignore the rich symmetries in the problem setting and, as a consequence, are sample-inefficient. This work identifies and utilizes the bi-equivariant symmetry within Keyframe IL to design a policy that generalizes to transformations of both the workspace and the objects grasped by the gripper. We make two main contributions: First, we analyze the bi-equivariance properties of the keyframe action scheme and propose a Keyframe Transporter derived from the Transporter Networks, which evaluates actions using cross-correlation between the features of the grasped object and the features of the scene. Second, we propose a computationally efficient coarse-to-fine SE(3) action evaluation scheme for reasoning the intertwined translation and rotation action. The resulting method outperforms strong Keyframe IL baselines by an average of >10% on a wide range of simulation tasks, and by an average of 55% in 4 physical experiments.

Ondrej Biza, Thomas Weng, Lingfeng Sun et al.

Reinforcement Learning (RL) has the potential to enable robots to learn from their own actions in the real world. Unfortunately, RL can be prohibitively expensive, in terms of on-robot runtime, due to inefficient exploration when learning from a sparse reward signal. Designing dense reward functions is labour-intensive and requires domain expertise. In our work, we propose GCR (Goal-Contrastive Rewards), a dense reward function learning method that can be trained on passive video demonstrations. By using videos without actions, our method is easier to scale, as we can use arbitrary videos. GCR combines two loss functions, an implicit value loss function that models how the reward increases when traversing a successful trajectory, and a goal-contrastive loss that discriminates between successful and failed trajectories. We perform experiments in simulated manipulation environments across RoboMimic and MimicGen tasks, as well as in the real world using a Franka arm and a Spot quadruped. We find that GCR leads to a more-sample efficient RL, enabling model-free RL to solve about twice as many tasks as our baseline reward learning methods. We also demonstrate positive cross-embodiment transfer from videos of people and of other robots performing a task. Website: https://gcr-robot.github.io/.

Haibo Zhao, Yu Qi, Boce Hu et al.

We present Generalizable Hierarchical Skill Learning (GSL), a novel framework for hierarchical policy learning that significantly improves policy generalization and sample efficiency in robot manipulation. One core idea of GSL is to use object-centric skills as an interface that bridges the high-level vision-language model and the low-level visual-motor policy. Specifically, GSL decomposes demonstrations into transferable and object-canonicalized skill primitives using foundation models, ensuring efficient low-level skill learning in the object frame. At test time, the skill-object pairs predicted by the high-level agent are fed to the low-level module, where the inferred canonical actions are mapped back to the world frame for execution. This structured yet flexible design leads to substantial improvements in sample efficiency and generalization of our method across unseen spatial arrangements, object appearances, and task compositions. In simulation, GSL trained with only 3 demonstrations per task outperforms baselines trained with 30 times more data by 15.5 percent on unseen tasks. In real-world experiments, GSL also surpasses the baseline trained with 10 times more data.

Owen Lewis Howell, Linfeng Zhao, Xupeng Zhu et al.

The global attention mechanism is one of the keys to the success of transformer architecture, but it incurs quadratic computational costs in relation to the number of tokens. On the other hand, equivariant models, which leverage the underlying geometric structures of problem instance, often achieve superior accuracy in physical, biochemical, computer vision, and robotic tasks, at the cost of additional compute requirements. As a result, existing equivariant transformers only support low-order equivariant features and local context windows, limiting their expressiveness and performance. This work proposes Clebsch-Gordan Transformer, achieving efficient global attention by a novel Clebsch-Gordon Convolution on $\SO(3)$ irreducible representations. Our method enables equivariant modeling of features at all orders while achieving ${O}(N \log N)$ input token complexity. Additionally, the proposed method scales well with high-order irreducible features, by exploiting the sparsity of the Clebsch-Gordon matrix. Lastly, we also incorporate optional token permutation equivariance through either weight sharing or data augmentation. We benchmark our method on a diverse set of benchmarks including n-body simulation, QM9, ModelNet point cloud classification and a robotic grasping dataset, showing clear gains over existing equivariant transformers in GPU memory size, speed, and accuracy.

Arsh Tangri, Nichols Crawford Taylor, Haojie Huang et al.

Contrastive Reinforcement Learning (CRL) provides a promising framework for extracting useful structured representations from unlabeled interactions. By pulling together state-action pairs and their corresponding future states, while pushing apart negative pairs, CRL enables learning nontrivial policies without manually designed rewards. In this work, we propose Equivariant CRL (ECRL), which further structures the latent space using equivariant constraints. By leveraging inherent symmetries in goal-conditioned manipulation tasks, our method improves both sample efficiency and spatial generalization. Specifically, we formally define Goal-Conditioned Group-Invariant MDPs to characterize rotation-symmetric robotic manipulation tasks, and build on this by introducing a novel rotation-invariant critic representation paired with a rotation-equivariant actor for Contrastive RL. Our approach consistently outperforms strong baselines across a range of simulated tasks in both state-based and image-based settings. Finally, we extend our method to the offline RL setting, demonstrating its effectiveness across multiple tasks.

Arsh Tangri, Ondrej Biza, Dian Wang et al.

Sample efficiency is critical when applying learning-based methods to robotic manipulation due to the high cost of collecting expert demonstrations and the challenges of on-robot policy learning through online Reinforcement Learning (RL). Offline RL addresses this issue by enabling policy learning from an offline dataset collected using any behavioral policy, regardless of its quality. However, recent advancements in offline RL have predominantly focused on learning from large datasets. Given that many robotic manipulation tasks can be formulated as rotation-symmetric problems, we investigate the use of $SO(2)$-equivariant neural networks for offline RL with a limited number of demonstrations. Our experimental results show that equivariant versions of Conservative Q-Learning (CQL) and Implicit Q-Learning (IQL) outperform their non-equivariant counterparts. We provide empirical evidence demonstrating how equivariance improves offline learning algorithms in the low-data regime.

Haojie Huang, Karl Schmeckpeper, Dian Wang et al.

Humans can imagine goal states during planning and perform actions to match those goals. In this work, we propose Imagination Policy, a novel multi-task key-frame policy network for solving high-precision pick and place tasks. Instead of learning actions directly, Imagination Policy generates point clouds to imagine desired states which are then translated to actions using rigid action estimation. This transforms action inference into a local generative task. We leverage pick and place symmetries underlying the tasks in the generation process and achieve extremely high sample efficiency and generalizability to unseen configurations. Finally, we demonstrate state-of-the-art performance across various tasks on the RLbench benchmark compared with several strong baselines and validate our approach on a real robot.

Robert Platt, Rossella Arcucci, Cédric M. John

Hyperspectral data acquired by the Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) have allowed for unparalleled mapping of the surface mineralogy of Mars. Due to sensor degradation over time, a significant portion of the recently acquired data is considered unusable. Here a new data-driven model architecture, Noise2Noise4Mars (N2N4M), is introduced to remove noise from CRISM images. Our model is self-supervised and does not require zero-noise target data, making it well suited for use in Planetary Science applications where high quality labelled data is scarce. We demonstrate its strong performance on synthetic-noise data and CRISM images, and its impact on downstream classification performance, outperforming benchmark methods on most metrics. This allows for detailed analysis for critical sites of interest on the Martian surface, including proposed lander sites.

Xupeng Zhu, Dian Wang, Ondrej Biza et al.

In planar grasp detection, the goal is to learn a function from an image of a scene onto a set of feasible grasp poses in $\mathrm{SE}(2)$. In this paper, we recognize that the optimal grasp function is $\mathrm{SE}(2)$-equivariant and can be modeled using an equivariant convolutional neural network. As a result, we are able to significantly improve the sample efficiency of grasp learning, obtaining a good approximation of the grasp function after only 600 grasp attempts. This is few enough that we can learn to grasp completely on a physical robot in about 1.5 hours.

Haojie Huang, Dian Wang, Robin Walters et al.

Transporter Net is a recently proposed framework for pick and place that is able to learn good manipulation policies from a very few expert demonstrations. A key reason why Transporter Net is so sample efficient is that the model incorporates rotational equivariance into the pick module, i.e. the model immediately generalizes learned pick knowledge to objects presented in different orientations. This paper proposes a novel version of Transporter Net that is equivariant to both pick and place orientation. As a result, our model immediately generalizes place knowledge to different place orientations in addition to generalizing pick knowledge as before. Ultimately, our new model is more sample efficient and achieves better pick and place success rates than the baseline Transporter Net model.

Ondrej Biza, Thomas Kipf, David Klee et al.

World models for environments with many objects face a combinatorial explosion of states: as the number of objects increases, the number of possible arrangements grows exponentially. In this paper, we learn to generalize over robotic pick-and-place tasks using object-factored world models, which combat the combinatorial explosion by ensuring that predictions are equivariant to permutations of objects. Previous object-factored models were limited either by their inability to model actions, or by their inability to plan for complex manipulation tasks. We build on recent contrastive methods for training object-factored world models, which we extend to model continuous robot actions and to accurately predict the physics of robotic pick-and-place. To do so, we use a residual stack of graph neural networks that receive action information at multiple levels in both their node and edge neural networks. Crucially, our learned model can make predictions about tasks not represented in the training data. That is, we demonstrate successful zero-shot generalization to novel tasks, with only a minor decrease in model performance. Moreover, we show that an ensemble of our models can be used to plan for tasks involving up to 12 pick and place actions using heuristic search. We also demonstrate transfer to a physical robot.

Haojie Huang, Ziyi Yang, Robert Platt

Shape completion, the problem of inferring the complete geometry of an object given a partial point cloud, is an important problem in robotics and computer vision. This paper proposes the Graph Attention Shape Completion Network (GASCN), a novel neural network model that solves this problem. This model combines a graph-based model for encoding local point cloud information with an MLP-based architecture for encoding global information. For each completed point, our model infers the normal and extent of the local surface patch which is used to produce dense yet precise shape completions. We report experiments that demonstrate that GASCN outperforms standard shape completion methods on a standard benchmark drawn from the Shapenet dataset.

Dian Wang, Robin Walters, Xupeng Zhu et al.

Recently, a variety of new equivariant neural network model architectures have been proposed that generalize better over rotational and reflectional symmetries than standard models. These models are relevant to robotics because many robotics problems can be expressed in a rotationally symmetric way. This paper focuses on equivariance over a visual state space and a spatial action space -- the setting where the robot action space includes a subset of $\rm{SE}(2)$. In this situation, we know a priori that rotations and translations in the state image should result in the same rotations and translations in the spatial action dimensions of the optimal policy. Therefore, we can use equivariant model architectures to make $Q$ learning more sample efficient. This paper identifies when the optimal $Q$ function is equivariant and proposes $Q$ network architectures for this setting. We show experimentally that this approach outperforms standard methods in a set of challenging manipulation problems.

Tarik Kelestemur, Colin Keil, John P. Whitney et al.

Localizing and tracking the pose of robotic grippers are necessary skills for manipulation tasks. However, the manipulators with imprecise kinematic models (e.g. low-cost arms) or manipulators with unknown world coordinates (e.g. poor camera-arm calibration) cannot locate the gripper with respect to the world. In these circumstances, we can leverage tactile feedback between the gripper and the environment. In this paper, we present learnable Bayes filter models that can localize robotic grippers using tactile feedback. We propose a novel observation model that conditions the tactile feedback on visual maps of the environment along with a motion model to recursively estimate the gripper's location. Our models are trained in simulation with self-supervision and transferred to the real world. Our method is evaluated on a tabletop localization task in which the gripper interacts with objects. We report results in simulation and on a real robot, generalizing over different sizes, shapes, and configurations of the objects.

Hai Nguyen, Brett Daley, Xinchao Song et al.

Many important robotics problems are partially observable in the sense that a single visual or force-feedback measurement is insufficient to reconstruct the state. Standard approaches involve learning a policy over beliefs or observation-action histories. However, both of these have drawbacks; it is expensive to track the belief online, and it is hard to learn policies directly over histories. We propose a method for policy learning under partial observability called the Belief-Grounded Network (BGN) in which an auxiliary belief-reconstruction loss incentivizes a neural network to concisely summarize its input history. Since the resulting policy is a function of the history rather than the belief, it can be executed easily at runtime. We compare BGN against several baselines on classic benchmark tasks as well as three novel robotic touch-sensing tasks. BGN outperforms all other tested methods and its learned policies work well when transferred onto a physical robot.

Marcus Gualtieri, Robert Platt

We consider robotic pick-and-place of partially visible, novel objects, where goal placements are non-trivial, e.g., tightly packed into a bin. One approach is (a) use object instance segmentation and shape completion to model the objects and (b) use a regrasp planner to decide grasps and places displacing the models to their goals. However, it is critical for the planner to account for uncertainty in the perceived models, as object geometries in unobserved areas are just guesses. We account for perceptual uncertainty by incorporating it into the regrasp planner's cost function. We compare seven different costs. One of these, which uses neural networks to estimate probability of grasp and place stability, consistently outperforms uncertainty-unaware costs and evaluates faster than Monte Carlo sampling. On a real robot, the proposed cost results in successfully packing objects tightly into a bin 7.8% more often versus the commonly used minimum-number-of-grasps cost.

Dian Wang, Colin Kohler, Robert Platt

In the spatial action representation, the action space spans the space of target poses for robot motion commands, i.e. SE(2) or SE(3). This approach has been used to solve challenging robotic manipulation problems and shows promise. However, the method is often limited to a three dimensional action space and short horizon tasks. This paper proposes ASRSE3, a new method for handling higher dimensional spatial action spaces that transforms an original MDP with high dimensional action space into a new MDP with reduced action space and augmented state space. We also propose SDQfD, a variation of DQfD designed for large action spaces. ASRSE3 and SDQfD are evaluated in the context of a set of challenging block construction tasks. We show that both methods outperform standard baselines and can be used in practice on real robotics systems.

Ulrich Viereck, Kate Saenko, Robert Platt

Learning control policies for visual servoing in novel environments is an important problem. However, standard model-free policy learning methods are slow. This paper explores planner cloning: using behavior cloning to learn policies that mimic the behavior of a full-state motion planner in simulation. We propose Penalized Q Cloning (PQC), a new behavior cloning algorithm. We show that it outperforms several baselines and ablations on some challenging problems involving visual servoing in novel environments while avoiding obstacles. Finally, we demonstrate that these policies can be transferred effectively onto a real robotic platform, achieving approximately an 87% success rate both in simulation and on a real robot.

Marcus Gualtieri, Robert Platt

Learning generalizable skills in robotic manipulation has long been challenging due to real-world sized observation and action spaces. One method for addressing this problem is attention focus -- the robot learns where to attend its sensors and irrelevant details are ignored. However, these methods have largely not caught on due to the difficulty of learning a good attention policy and the added partial observability induced by a narrowed window of focus. This article addresses the first issue by constraining gazes to a spatial hierarchy. For the second issue, we identify a case where the partial observability induced by attention does not prevent Q-learning from finding an optimal policy. We conclude with real-robot experiments on challenging pick-place tasks demonstrating the applicability of the approach.

Brayan S. Zapata-Impata, Vikrant Shah, Hanumant Singh et al.

Tasks in outdoor open world environments are now ripe for automation with mobile manipulators. The dynamic, unstructured and unknown environments associated with such tasks -- a prime example would be collecting roadside trash -- makes them particularly challenging. In this paper we present an approach to solving the problem of picking up, transporting, and dropping off novel objects outdoors. Our solution integrates a navigation system, a grasp detection and planning system, and a custom task planner. We perform experiments that demonstrate that the system can be used to transport a wide class of novel objects (trash bags, general garbage, gardening tools and fruits) in unstructured settings outdoors with a relatively high end-to-end success rate of 85%. See it at work at: https://youtu.be/93nWXhaGEWA

Dian Wang, Colin Kohler, Andreas ten Pas et al.

Assistive robot manipulators must be able to autonomously pick and place a wide range of novel objects to be truly useful. However, current assistive robots lack this capability. Additionally, assistive systems need to have an interface that is easy to learn, to use, and to understand. This paper takes a step forward in this direction. We present a robot system comprised of a robotic arm and a mobility scooter that provides both pick-and-drop and pick-and-place functionality for open world environments without modeling the objects or environment. The system uses a laser pointer to directly select an object in the world, with feedback to the user via projecting an interface into the world. Our evaluation over several experimental scenarios shows a significant improvement in both runtime and grasp success rate relative to a baseline from the literature [5], and furthermore demonstrates accurate pick and place capabilities for tabletop scenarios.

Ulrich Viereck, Xingchao Peng, Kate Saenko et al.

This paper proposes an approach to domain transfer based on a pairwise loss function that helps transfer control policies learned in simulation onto a real robot. We explore the idea in the context of a 'category level' manipulation task where a control policy is learned that enables a robot to perform a mating task involving novel objects. We explore the case where depth images are used as the main form of sensor input. Our experimental results demonstrate that proposed method consistently outperforms baseline methods that train only in simulation or that combine real and simulated data in a naive way.

Robert Platt, Colin Kohler, Marcus Gualtieri

In applications of deep reinforcement learning to robotics, it is often the case that we want to learn pose invariant policies: policies that are invariant to changes in the position and orientation of objects in the world. For example, consider a peg-in-hole insertion task. If the agent learns to insert a peg into one hole, we would like that policy to generalize to holes presented in different poses. Unfortunately, this is a challenge using conventional methods. This paper proposes a novel state and action abstraction that is invariant to pose shifts called \textit{deictic image maps} that can be used with deep reinforcement learning. We provide broad conditions under which optimal abstract policies are optimal for the underlying system. Finally, we show that the method can help solve challenging robotic manipulation problems.

Marcus Gualtieri, Robert Platt

We address a class of manipulation problems where the robot perceives the scene with a depth sensor and can move its end effector in a space with six degrees of freedom -- 3D position and orientation. Our approach is to formulate the problem as a Markov decision process (MDP) with abstract yet generally applicable state and action representations. Finding a good solution to the MDP requires adding constraints on the allowed actions. We develop a specific set of constraints called hierarchical $\text{SE}(3)$ sampling (HSE3S) which causes the robot to learn a sequence of gazes to focus attention on the task-relevant parts of the scene. We demonstrate the effectiveness of our approach on three challenging pick-place tasks (with novel objects in clutter and nontrivial places) both in simulation and on a real robot, even though all training is done in simulation.

Andrew Levy, Robert Platt, Kate Saenko

Reinforcement Learning (RL) algorithms can suffer from poor sample efficiency when rewards are delayed and sparse. We introduce a solution that enables agents to learn temporally extended actions at multiple levels of abstraction in a sample efficient and automated fashion. Our approach combines universal value functions and hindsight learning, allowing agents to learn policies belonging to different time scales in parallel. We show that our method significantly accelerates learning in a variety of discrete and continuous tasks.

Andrew Levy, George Konidaris, Robert Platt et al.

Hierarchical agents have the potential to solve sequential decision making tasks with greater sample efficiency than their non-hierarchical counterparts because hierarchical agents can break down tasks into sets of subtasks that only require short sequences of decisions. In order to realize this potential of faster learning, hierarchical agents need to be able to learn their multiple levels of policies in parallel so these simpler subproblems can be solved simultaneously. Yet, learning multiple levels of policies in parallel is hard because it is inherently unstable: changes in a policy at one level of the hierarchy may cause changes in the transition and reward functions at higher levels in the hierarchy, making it difficult to jointly learn multiple levels of policies. In this paper, we introduce a new Hierarchical Reinforcement Learning (HRL) framework, Hierarchical Actor-Critic (HAC), that can overcome the instability issues that arise when agents try to jointly learn multiple levels of policies. The main idea behind HAC is to train each level of the hierarchy independently of the lower levels by training each level as if the lower level policies are already optimal. We demonstrate experimentally in both grid world and simulated robotics domains that our approach can significantly accelerate learning relative to other non-hierarchical and hierarchical methods. Indeed, our framework is the first to successfully learn 3-level hierarchies in parallel in tasks with continuous state and action spaces.

Marcus Gualtieri, Andreas ten Pas, Robert Platt

We propose a novel formulation of robotic pick and place as a deep reinforcement learning (RL) problem. Whereas most deep RL approaches to robotic manipulation frame the problem in terms of low level states and actions, we propose a more abstract formulation. In this formulation, actions are target reach poses for the hand and states are a history of such reaches. We show this approach can solve a challenging class of pick-place and regrasping problems where the exact geometry of the objects to be handled is unknown. The only information our method requires is: 1) the sensor perception available to the robot at test time; 2) prior knowledge of the general class of objects for which the system was trained. We evaluate our method using objects belonging to two different categories, mugs and bottles, both in simulation and on real hardware. Results show a major improvement relative to a shape primitives baseline.

Andreas ten Pas, Marcus Gualtieri, Kate Saenko et al.

Recently, a number of grasp detection methods have been proposed that can be used to localize robotic grasp configurations directly from sensor data without estimating object pose. The underlying idea is to treat grasp perception analogously to object detection in computer vision. These methods take as input a noisy and partially occluded RGBD image or point cloud and produce as output pose estimates of viable grasps, without assuming a known CAD model of the object. Although these methods generalize grasp knowledge to new objects well, they have not yet been demonstrated to be reliable enough for wide use. Many grasp detection methods achieve grasp success rates (grasp successes as a fraction of the total number of grasp attempts) between 75% and 95% for novel objects presented in isolation or in light clutter. Not only are these success rates too low for practical grasping applications, but the light clutter scenarios that are evaluated often do not reflect the realities of real world grasping. This paper proposes a number of innovations that together result in a significant improvement in grasp detection performance. The specific improvement in performance due to each of our contributions is quantitatively measured either in simulation or on robotic hardware. Ultimately, we report a series of robotic experiments that average a 93% end-to-end grasp success rate for novel objects presented in dense clutter.

Ulrich Viereck, Andreas ten Pas, Kate Saenko et al.

We want to build robots that are useful in unstructured real world applications, such as doing work in the household. Grasping in particular is an important skill in this domain, yet it remains a challenge. One of the key hurdles is handling unexpected changes or motion in the objects being grasped and kinematic noise or other errors in the robot. This paper proposes an approach to learning a closed-loop controller for robotic grasping that dynamically guides the gripper to the object. We use a wrist-mounted sensor to acquire depth images in front of the gripper and train a convolutional neural network to learn a distance function to true grasps for grasp configurations over an image. The training sensor data is generated in simulation, a major advantage over previous work that uses real robot experience, which is costly to obtain. Despite being trained in simulation, our approach works well on real noisy sensor images. We compare our controller in simulated and real robot experiments to a strong baseline for grasp pose detection, and find that our approach significantly outperforms the baseline in the presence of kinematic noise, perceptual errors and disturbances of the object during grasping.

Marcus Gualtieri, James Kuczynski, Abraham M. Shultz et al.

Many people with motor disabilities are unable to complete activities of daily living (ADLs) without assistance. This paper describes a complete robotic system developed to provide mobile grasping assistance for ADLs. The system is comprised of a robot arm from a Rethink Robotics Baxter robot mounted to an assistive mobility device, a control system for that arm, and a user interface with a variety of access methods for selecting desired objects. The system uses grasp detection to allow previously unseen objects to be picked up by the system. The grasp detection algorithms also allow for objects to be grasped in cluttered environments. We evaluate our system in a number of experiments on a large variety of objects. Overall, we achieve an object selection success rate of 88% and a grasp detection success rate of 90% in a non-mobile scenario, and success rates of 89% and 72% in a mobile scenario.