When moving objects that are too bulky or heavy to be grasped or lifted, robotic manipulation can benefit from the object's interaction with the support surface and its natural dynamics under gravity. In this work, we show that such dynamic, underactuated manipulation capability can be acquired through reinforcement learning and deployed on real robot systems. First, we present a framework to learn a control policy for object transport in a dynamic simulation environment, featuring the object and the support surface. We then demonstrate successful object locomotion with the learned policy through a set of simulated and real-world experiments, performed with a robot arm and an aerial robot interacting with the object in a non-prehensile manner. While the object, which is in contact with the support surface, oscillates sideways passively under gravity, the robot uses the learned policy to move the object forward with a steady gait by regulating the mechanical energy and the posture of the object. Our experiment results show that the learned policy can transport the object through unmodeled effects of terrain and perturbation.

We present a data-driven approach for effective bin picking from clutter. Recent bin picking solutions usually lead to a direct pinch grasp on a target object without addressing any other potential contact interaction in clutter. However, appropriate physical interaction can be essential to successful singulation and subsequent secure picking, the goal of bin picking. In this work, we contribute a framework that learns physically interactive actions for object picking end-to-end from a visual input in a self-supervised manner. The learned actions enable the robot to purposefully interact with a target object by performing a digging operation through the clutter. By leveraging a fully convolutional network (FCN), we predict picking success probabilities for a set of interactive action primitives that will in turn specify an optimal action to perform. The FCN is trained in a simulated environment through trial and error. Moreover, new datasets are collected using the latest network through iterative self-supervision. Extensive real-world bin picking experiments show the effectiveness and generalizability of the approach.

Spatio-temporal information is key to resolve occlusion and depth ambiguity in 3D human pose estimation. Previous methods have focused on either temporal contexts or local-to-global architectures that embed fixed-length spatiotemporal information. To date, there have not been effective proposals to simultaneously and flexibly capture varying spatiotemporal sequences and effectively achieves real-time 3D human pose estimation. In this work, we improve the learning of kinematic constraints in the human skeleton: posture, local kinematic connections, and symmetry by modeling local and global spatial information via attention mechanisms. To adapt to single- and multi-frame estimation, the dilated temporal model is employed to process varying skeleton sequences. Also, importantly, we carefully design the interleaving of spatial semantics with temporal dependencies to achieve a synergistic effect. To this end, we propose a simple yet effective graph attention spatio-temporal convolutional network (GAST-Net) that comprises of interleaved temporal convolutional and graph attention blocks. Experiments on two challenging benchmark datasets (Human3.6M and HumanEva-I) and YouTube videos demonstrate that our approach effectively mitigates depth ambiguity and self-occlusion, generalizes to half upper body estimation, and achieves competitive performance on 2D-to-3D video pose estimation. Code, video, and supplementary information is available at: http://www.juanrojas.net/gast/

Event detection is a critical feature in data-driven systems as it assists with the identification of nominal and anomalous behavior. Event detection is increasingly relevant in robotics as robots operate with greater autonomy in increasingly unstructured environments. In this work, we present an accurate, robust, fast, and versatile measure for skill and anomaly identification. A theoretical proof establishes the link between the derivative of the log-likelihood of the HMM filtered belief state and the latest emission probabilities. The key insight is the inverse relationship in which gradient analysis is used for skill and anomaly identification. Our measure showed better performance across all metrics than all but one related state-of-the-art works. The result is broadly applicable to domains that use HMMs for event detection. Supplemental information, code, data, and videos can be found at [1].

Robots are increasingly entering uncertain and unstructured environments. Within these, robots are bound to face unexpected external disturbances like accidental human or tool collisions. Robots must develop the capacity to respond to unexpected events. That is not only identifying the sudden anomaly, but also deciding how to handle it. In this work, we contribute a recovery policy that allows a robot to recovery from various anomalous scenarios across different tasks and conditions in a consistent and robust fashion. The system organizes tasks as a sequence of nodes composed of internal modules such as motion generation and introspection. When an introspection module flags an anomaly, the recovery strategy is triggered and reverts the task execution by selecting a target node as a function of a state dependency chart. The new skill allows the robot to overcome the effects of the external disturbance and conclude the task. Our system recovers from accidental human and tool collisions in a number of tasks. Of particular importance is the fact that we test the robustness of the recovery system by triggering anomalies at each node in the task graph showing robust recovery everywhere in the task. We also trigger multiple and repeated anomalies at each of the nodes of the task showing that the recovery system can consistently recover anywhere in the presence of strong and pervasive anomalous conditions. Robust recovery systems will be key enablers for long-term autonomy in robot systems. Supplemental information including videos, code, and result analysis can be found at [1].

Robotic failure is all too common in unstructured robot tasks. Despite well-designed controllers, robots often fail due to unexpected events. Robots under a sense-plan-act paradigm do not have an additional loop to check their actions. In this work, we present a principled methodology to bootstrap online robot introspection for contact tasks. In effect, we seek to enable the robot to recognize and expect its behavior, else detect anomalies. We postulated that noisy wrench data inherently contains patterns that can be effectively represented by a vocabulary. The vocabulary is obtained by segmenting and encoding data. And when wrench information represents a sequence of sub-tasks, the vocabulary represents a set of words or sentence and provides a unique identifier. The grammar, which can also include unexpected events, was classified both offline and online for simulated and real robot experiments. Multi-class Support Vector Machines (SVMs) were used offline, while online probabilistic SVMs were used to give temporal confidence to the introspection result. Our work's contribution is the presentation of a generalizable online semantic scheme that enables a robot to understand its high-level state whether nominal or anomalous. It is shown to work in offline and online scenarios for a particularly challenging contact task: snap assemblies. We perform the snap assembly in one-arm simulated and real one-arm experiments and a simulated two-arm experiment. The data set itself is also fully available online and provides a valuable resource by itself for this type of contact task. Our verification mechanism can be used by high-level planners or reasoning systems to enable intelligent failure recovery or determine the next most optimal manipulation skill to be used. Supplemental information, code, data, and other supporting documentation can be found at [1].

Re-configurable modular robotic (RMR) systems are advantageous for their reconfigurability and versatility. A new modular robot can be built for a specific task by using modules as building blocks. However, constructing a kinematic model for a newly conceived robot requires significant work. Due to the finite size of module-types, models of all module-types can be built individually and stored in a database beforehand. With this priori knowledge, the model construction process can be automated by detecting the modules and their corresponding interconnections. Previous literature proposed theoretical frameworks for constructing kinematic models of modular robots, assuming that such information was known a priori. While well-devised mechanisms and built-in sensors can be employed to detect these parameters, they significantly complicate the module design and thus are expensive. In this paper, we propose a vision-based method to identify kinematic chains and automatically construct robot models for modular robots. Each module is affixed with augmented reality (AR) tags that are encoded with unique IDs. An image of a modular robot is taken and the detected modules are recognized by querying a database that maintains all module information. The poses of detected module-links are used to compute: (i) the connection between modules and (ii) joint angles of joint-modules. Finally, the robot serial-link chain is identified and the kinematic model is constructed and visualized. Our experimental results validate the effectiveness of our approach. While implementation with only our RMR is shown, our method can be applied to other RMRs where self-identification is not possible.

Failure detection and correction is essential in robust systems. In robotics, failure detection has focused on traditional parts assembly, tool breakage, and threaded fastener assembly. However, not much work has focused on sub-mode failure classification. This is an important step in order to provide accurate failure recovery. Our work implemented a novel failure characterization scheme for cantilever snap assemblies. The approach identified exemplars that characterized salient features for specific deviations from a nominal trajectory. Then, a rule based approach with statistical measures was used to identify failure and classify failure sub-modes. Failure sub-mode classification was evaluated by using a reliability measure. Our work classified failure deviations with 88% accuracy. Varying success was experienced in correlating failure deviation modes. Cases with only 1-deviation had 86% accuracy, cases with 2-deviations had 67% accuracy, and cases with 3 deviations had 55% accuracy. Our work is an important step in failure characterization of complex geometrical parts and serves as a stepping stone to enact failure recovery.

Snap assembly automation remains a challenging task. While progress is being made in localization of parts, force controllers, and control strategies, little work has been done to help the robot reason about its current state, such that if necessary, the robot can assume corrective actions to accomplish the task. Error prone situations caused by the unexpected motion of parts, localization errors, jamming or wedging, cannot be solved by force controllers alone. For this reason we propose a snap assemblies verification system for cantilever-snap fasteners. The verification works in concert with a control strategy that makes use of constraint designs embedded in the snap parts' physical design. The constrained assembly motion generates similar sensory-signal patterns across trials that facilitates force signal discrimination into higher level abstractions of intuitive behavior. This work's contribution is the design of a hierarchical taxonomy for cantilever-snap verification based on increasingly abstract layers that encode relative-change in the task's force signatures. A five-layered taxonomy is built on the concept that relative-change patterns can be classified through a small category set and aided by contextual information. The verification system yielded human apropos intuitive categorizations of task behavior for every state and effectively determined the assembly result. This simple yet effective approach will be expanded to perform probabilistic online system verification to aid in fault tolerance and the automation of cantilever-based snap assemblies.