ICRA 2018 Papers

This paper presents an autonomous robot system that is designed to autonomously search for and geo-localize potential underwater archaeological sites. The system, based on Autonomous Underwater Vehicles, invokes a multi-step pipeline. First, the AUV constructs a high altitude scan over a large area to collect low-resolution side scan sonar data. Second, image processing software is employed to automatically detect and identify potential sites of interest. Third, a ranking algorithm assigns importance scores to each site. Fourth, an AUV path planner is used to plan a time-limited path that visits sites with a high importance at a low altitude to acquire high-resolution sonar data. Last, the AUV is deployed to follow this path. This system was implemented and evaluated during an archaeological survey located along the coast of Malta. These experiments demonstrated that the system is able to identify valuable archaeological sites accurately and efficiently in a large previously unsurveyed area. Also, the planned missions led to the discovery of a historical plane wreck whose location was previously unknown.

The efficiency of sampling-based motion planning algorithms is dependent on how well a steering procedure is capable of capturing both system dynamics and configuration space geometry to connect sample configurations. This paper considers how metrics describing local system dynamics may be combined with convex subsets of the free space to describe the local behavior of a steering function for sampling-based planners. Subsequently, a framework for using these subsets to extend the steering procedure to incorporate this information is introduced. To demonstrate our framework, three specific metrics are considered: the LQR cost-to-go function, a Gram matrix derived from system linearization, and the Mahalanobis distance of a linear-Gaussian system. Finally, numerical tests are conducted for a second-order linear system, a kinematic unicycle, and a linear-Gaussian system to demonstrate that our framework increases the connectivity of sampling-based planners and allows them to better explore the free space.

Field multi-robot missions face numerous unavoidable disturbances, such as delays in executing tasks and intermittent communications. Coping with such disturbances requires to endow the robots with high-level decision skills. We present a distributed decision architecture based first on a hybrid planner that can manage decentralized repairs with partial communication, and secondly on a distributed execution algorithm that efficiently propagates delays. This architecture has been successfully experimented on the field for the achievement of surveillance missions involving eight (8) real autonomous aerial and ground robots.

At the moment, surgeons struggle curing a widespread eye disease known as retinal vein occlusion where clots obstruct the retinal vessels. Latter vascular disorder involves black spots in people's eyesight and lead eventually to blindness. A recent promising treatment consists in flushing a thrombolytic agent inside the clotted retinal vessels. The surgery implies puncturing vessels ranging from 50 to 400 microns diameter on the backside of the eye, namely the retina. Latest research succeeded in tackling several challenges around this operation: the surgeon's hand tremor and the high precision required amongst other requirements. Despite several breakthroughs, the surgeon only relies on a microscope to perform the surgery through the patient eye's lens, giving poor depth perception to properly puncture the retinal vessels. This way, the surgeon is most likely to pierce through the vessel and inject the thrombolytic drug under the retina, which would endanger the person's eyesight. In this paper, we investigate the use of a novel bio-impedance sensor developed for eye surgery. Together with this new sensor, a detection algorithm has been developed to detect the puncture and double puncture events to give a feedback to the operator of the system. As far as we are aware of, such technology doesn't exist yet in eye surgery to tackle the depth perception question. This paper aims at demonstrating the benefits of this technology.

Ahstract- This paper presents a method to estimate the original location and the mass of an instantaneous release of hazardous material into the atmosphere. It is formulated as an inverse problem, where concentration observations from a mobile sensor are fused with meteorological information and a Gaussian puff dispersion model to characterise the source. Bayes' theorem is used to estimate the parameters of the release taking into account the uncertainty that exists in the dispersion parameters and meteorological variables. An information based reward is used to guide an unmanned aerial vehicle equipped with a chemical sensor to the expected most informative measurement locations. Simulation results compare the performance between a single mobile sensor with various amounts of static sensors.

Users play an integral role in the performance of many robotic systems, and robotic systems must account for differences in users to improve collaborative performance. Much of the work in adapting to users has focused on designing teleoperation controllers that adjust to extrinsic user indicators such as force, or intent, but do not adjust to intrinsic user qualities. In contrast, the Human-Robot Interaction community has extensively studied intrinsic user qualities, but results may not rapidly be fed back into autonomy design. Here we provide foundational evidence for a new strategy that augments current shared control, and provide a mechanism to directly feed back results from the HRI community into autonomy design. Our evidence is based on a study examining the impact of the user quality “locus of control” on telepresence robot performance. Our results support our hypothesis that key user qualities can be inferred from human-robot interactions (such as through path deviation or time to completion) and that switching or adaptive autonomies might improve shared control performance.

Knowledge of the mass and inertial parameters of a humanoid robot is crucial for the development of model-based controller and motion planning in dynamics situation. Parameters are usually provided from Computer Aided Design (CAD) data and thus inaccurate specially if the robot is modified over time. In this paper, a practical method consisting of hanging a humanoid robot to a fix force sensor to perform its dynamic identification is proposed. This allows, contrary to the literature, to generate very exciting and dynamic motions to identify most of the elements of the inertia tensors in a reduced amount of time. This procedure transforms an instable floating base legged humanoid robot to a safe fix base tree structure robot which makes easier to generate optimal exciting motions. Because of a better excitation the overall trajectory lasts for less than a minute. The method was experimentally validated with a HOAP3 humanoid robot and using a 6-axis force sensor. A reduction of 3 times in average of the RMS difference between measured external reaction forces and moments and their estimates from CAD data was obtained with a single minute of optimal exciting motions.

Smart manufacturing technologies are emerging which combine industrial equipment with Internet-of-Things (IoT) sensors to monitor and improve productivity of manufacturing. This allows for new opportunities to explore algorithms for predicting machine failures from attached sensor data. This paper presents a solution to non-invasively upgrade an existing machine with an Inertial Machine Monitoring System (IMMS) to detect and classify equipment failure or degraded state. We also provide a strategy to optimize the amount, placement locations, and efficiency of the sensors. In experiments, the system collected data from 36 inertial sensors placed at multiple locations on a 3D printer. Normal operation vs. 10 types of realworld abnormal equipment behavior (loose belt, failures of machine components) were detected and classified by Support Vector Machines and Neural Networks. Using under 1 minute of recording while running a test print, a recursively discovered best subset of 4 to 9 sensors yielded 11-way classification accuracy over 99%. Our results suggest that even a small sensor network and short test program can yield effective detection of machine degraded state and can facilitate early remediation.

Movement Primitives are a well studied and widely applied concept in modern robotics. Composing primitives out of an existing library, however, has shown to be a challenging problem. We propose the use of probabilistic context-free grammars to sequence a series of primitives to generate complex robot policies from a given library of primitives. The rule-based nature of formal grammars allows an intuitive encoding of hierarchically and recursively structured tasks. This hierarchical concept strongly connects with the way robot policies can be learned, organized, and re-used. However, the induction of context-free grammars has proven to be a complicated and yet unsolved challenge. In this work, we exploit the physical nature of robot movement primitives to restrict and efficiently search the grammar space. The grammar is learned applying a Markov Chain Monte Carlo optimization over the posteriors of the grammars given the observations. The proposal distribution is defined as a mixture over the probabilities of the operators connecting the search space. Restrictions to these operators guarantee continuous sequences while reducing the grammar space. We validate our method on a redundant 7 degree-of-freedom lightweight robotic arm on tasks that require the generation of complex sequences consisting of simple movement primitives.

Coverage Path Planning (CPP) describes the process of generating robot trajectories that fully cover an area or volume. Applications are, amongst many others, mobile cleaning robots, lawn mowing robots or harvesting machines in agriculture. Many approaches and facets of this problem have been discussed in literature but despite the availability of several surveys on the topic there is little work on quantitative assessment and comparison of different coverage path planning algorithms. This paper analyzes six popular off-line coverage path planning methods, applicable to previously recorded maps, in the setting of indoor coverage path planning on room-sized units. The implemented algorithms are thoroughly compared on a large dataset of over 550 rooms with and without furniture.

In realistic environments, fully specifying a task model such that a robot can perform a task in all situations is impractical. In this work, we present Incremental Task Modification via Corrective Demonstrations (ITMCD), a novel algorithm that allows a robot to update a learned model by making use of corrective demonstrations from an end-user in its environment. We propose three different types of model updates that make structural changes to a finite state automaton (FSA) representation of the task by first converting the FSA into a state transition auto-regressive hidden Markov model (STARHMM). The STARHMM's probabilistic properties are then used to perform approximate Bayesian model selection to choose the best model update, if any. We evaluate ITMCD Model Selection in a simulated block sorting domain and the full algorithm on a real-world pouring task. The simulation results show our approach can choose new task models that sufficiently incorporate new demonstrations while remaining as simple as possible. The results from the pouring task show that ITMCD performs well when the modeled segments of the corrective demonstrations closely comply with the original task model.

Continuous appearance shifts such as changes in weather and lighting conditions can impact the performance of deployed machine learning models. While unsupervised domain adaptation aims to address this challenge, current approaches do not utilise the continuity of the occurring shifts. In particular, many robotics applications exhibit these conditions and thus facilitate the potential to incrementally adapt a learnt model over minor shifts which integrate to massive differences over time. Our work presents an adversarial approach for lifelong, incremental domain adaptation which benefits from unsupervised alignment to a series of intermediate domains which successively diverge from the labelled source domain. We empirically demonstrate that our incremental approach improves handling of large appearance changes, e.g. day to night, on a traversable-path segmentation task compared with a direct, single alignment step approach. Furthermore, by approximating the feature distribution for the source domain with a generative adversarial network, the deployment module can be rendered fully independent of retaining potentially large amounts of the related source training data for only a minor reduction in performance.

In most complex missions, unexpected situations arise that may interfere with the planned execution of mission tasks. These situations result in the generation of contingency tasks that need to be executed before the originally planned tasks are completed. Potential contingency tasks may not always affect mission tasks due to the inherent uncertainty in the environment. Deferring action on a potential contingency task may incur a penalty in terms of wasted time due to idle robots if the contingency task becomes a bottleneck in the future. On the other hand, immediate action on a potential contingency task may incur a penalty in terms of wasted time if the contingency task did not actually impact the mission. When a contingency task is reported, the planner generates an updated plan that minimizes expected mission completion time by taking into account the probability of the contingency task impacting mission tasks, its effect on the mission, and its spatial location. We have characterized the performance of the algorithm through simulation experiments. We show that the proactive approach to contingency task management outperforms a conservative approach.

For the state-of-the-art of soft robotics, the current actuation mechanisms cannot produce shear forces, neither are the current stiffening mechanisms adaptive to various deformations. Consequently, the soft robots gain strength at the price of losing flexibility. To fill this gap, we proposed a new mechanism based on the muscle arrangements and incompressible property identified in biological hydrostatic skeletons. Beside longitudinal and transverse muscles, the proposed mechanism includes the oblique arrangement which is proved to play an indispensable role of producing shear forces. The effectiveness of the new mechanism is demonstrated through a benchmark problem - carrying a distributed load at the initial horizontal configuration, thus indicating an improved direction to realise shape-independent load-carrying capability of soft robotics. Furthermore, the proposed mechanism may explain how elephants coordinate the two contradicting properties, strength and flexibility, during their trunk manipulations.

When walking at high speeds, the swing legs of robots produce a non-negligible angular momentum rate. To accommodate this, we provide a reference trajectory generator for bipedal walking that incorporates predicted centroidal angular momentum at the planning stage. This can be done efficiently as the Centroidal Moment Pivot (CMP), Instantaneous Capture Point (ICP) and the center of mass (CoM) all have closed-form trajectory solutions due to their linear dynamics. This is then used to produce smooth, continuous trajectories. We furthermore provide a lightweight model to estimate angular momentum as induced during leg swing of the gait cycle. Our proposed trajectory generator is tested thoroughly in simulation and has been shown to successfully operate on the real hardware.

Inspired by the inchworm locomotion mechanism, this paper presents our recently developed self-deformable capsule-like robot. The robot has the actuated deformation capability that relies on a novel rigid elements-based morphing structure (REMS) and its soft actuation mechanisms. When the robot deforms, it generates the crawling locomotion behavior and thus friction waves between the robot and contact surface to facilitate the inchworm-like crawling movement. The paper starts reviewing the deformable properties of natural biological entities like capsules, presents state of the art of the current capsule-like robots, and details the bio-inspired design of the self-deformable capsule-like robot by describing the model of robot kinematics and its locomotion mechanism. Both simulation and experimental results validate the excellent performance of this capsule-like robot. The developed self-deformable capsule-like robot has the advantage of crawling on varied surfaces and it also has the capabilities to crawl in a variety of narrow pipes based on the deformation elicited locomotion nature of the robot.

This paper proposes an object modeling and grasping pipeline for humanoid robots. This work improves our previous approach based on superquadric functions. In particular, we speed up and refine the modeling process by using prior information on the object shape provided by an object classifier. We use our previous method for the computation of grasping pose to obtain pose candidates for both the robot hands and, then, we automatically choose the best candidate for grasping the object according to a given quality index. The performance of our pipeline has been assessed on a real robotic system, the iCub humanoid robot. The robot can grasp 18 objects of the YCB and iCub World datasets considerably different in terms of shape and dimensions with a high success rate.

Balance control to recover from a wide range of disturbances is an important skill for humanoid robots. Traditionally, researchers have often designed a balance controller by applying optimal control theory on a simplified model that abstracts the full-body dynamics. However, the resulting controller may not be able to recover from unexpected scenarios such as non-planar pushes, or fail to exploit full-body actions such as balancing with arm movements. This paper presents a learning framework for enhancing the performance of a model-based optimal controller by expanding the region of attraction (RoA). We train a control policy that generates additional control signals on top of the model-based controller using deep reinforcement learning techniques. Instead of relying on standard reinforcement learning formulations, we explicitly model the region of attraction and continuously adjust it during the training. By drawing the training disturbances at the boundary of the RoA, we can effectively expand the RoA while avoiding local minima. We test our learning framework for in-place balancing as well as balancing with stepping on a humanoid model in simulation.

This work proposes a process for efficiently searching over combinations of individual object 6D pose hypotheses in cluttered scenes, especially in cases involving occlusions and objects resting on each other. The initial set of candidate object poses is generated from state-of-the-art object detection and global point cloud registration techniques. The best scored pose per object by using these techniques may not be accurate due to overlaps and occlusions. Nevertheless, experimental indications provided in this work show that object poses with lower ranks may be closer to the real poses than ones with high ranks according to registration techniques. This motivates a global optimization process for improving these poses by taking into account scene-level physical interactions between objects. It also implies that the Cartesian product of candidate poses for interacting objects must be searched so as to identify the best scene-level hypothesis. To perform the search efficiently, the candidate poses for each object are clustered so as to reduce their number but still keep a sufficient diversity. Then, searching over the combinations of candidate object poses is performed through a Monte Carlo Tree Search (MCTS) process that uses the similarity between the observed depth image of the scene and a rendering of the scene given the hypothesized pose as a score that guides the search procedure. MCTS handles in a principled way the tradeoff between fine-tuning the most promising poses and exploring new ones, by using the Upper Confidence Bound (UCB) technique. Experimental results indicate that this process is able to quickly identify in cluttered scenes physically-consistent object poses that are significantly closer to ground truth compared to poses found by point cloud registration methods.