ICRA 2016 论文列表

There is growing interest in applying machine learning to motion planning. Potential applications are predicting an initial seed for trajectory optimization, predicting an effective heuristic for search based planning, and even predicting a planning algorithm for adaptive motion planning systems. In these situations, providing only a single prediction is unsatisfactory. It leads to many scenarios where the prediction suffers a high loss. In this paper, we advocate list prediction. Each predictor in a list focusses on different regions in the space of environments. This overcomes the shortcoming of a single predictor, and improves overall performance. A framework for list prediction, ConseqOpt, already exists. Our contribution is an extensive domain-specific treatment. We provide a rigorous and clear exposition of the procedure for training a list of predictors. We provide experimental results on a spectrum of motion planning applications. Each application contributes to understanding the behavior of list prediction. We observe that the benefit of list prediction over a single prediction is significant, irrespective of the application.

One of the key problems in planning and control of redundant robots is the fast generation of controls when multiple tasks and constraints need to be satisfied. In the literature, this problem is classically solved by multi-task prioritized approaches, where the priority of each task is determined by a weight function, describing the task strict/soft priority. In this paper, we propose to leverage machine learning techniques to learn the temporal profiles of the task priorities, represented as parametrized weight functions: we automatically determine their parameters through a stochastic optimization procedure. We show the effectiveness of the proposed method on a simulated 7 DOF Kuka LWR and both a simulated and a real Kinova Jaco arm. We compare the performance of our approach to a state-of-the-art method based on soft task prioritization, where the task weights are typically hand-tuned.

We present an approach to learn optimal navigation actions for assistance tasks in which the robot aims at efficiently reaching the final navigation goal of a human where service has to be provided. Always following the human at a close distance might hereby result in inefficient trajectories, since people regularly do not move on the shortest path to their destination (e.g., they move to grab the phone or make a note). Therefore, a service robot should infer the human's intended navigation goal and compute its own motion based on that prediction. We developed an approach that applies reinforcement learning to get a Q-function that determines for each pair of the robot's and human's relative positions the best navigation action for the robot. Our approach applies a prediction of the human's motion based on a softened Markov decision process (MDP). This MDP is independent from the navigation learning framework and is learned beforehand on previously observed trajectories. We thoroughly evaluated our method in simulation and on a real robot. As the experimental results show, our approach leads to foresighted navigation behavior and significantly reduces the path length and completion time compared to naive following strategies.

Imitation learning of tasks in multi-component robotic systems requires capturing concurrency and synchronization requirements in addition to task structure. Learning time-critical tasks depends furthermore on the ability to model temporal elements in demonstrations. This paper proposes a modeling framework based on Petri nets capable of modeling these aspects in a programming by demonstration context. In the proposed approach, models of tasks are constructed from segmented demonstrations as task Petri nets, which can be executed as discrete controllers for reproduction. We present algorithms that automatically construct models from demonstrations, showing how elements of time-critical tasks can be mapped into task Petri net elements. The approach is validated by an experiment in which a robot plays a musical passage on a keyboard.

This paper presents a new approach for fast collision detection in high dimensional configuration spaces for Rapidly-exploring Random Trees (RRT) motion planning. The proposed method is based upon Gaussian Mixture Models (GMM) that are learned using an incremental Expectation Maximization clustering algorithm trained online using exemplars provided by a slow, conventional kinematic-based collision detection routine. The number of collision checks needed can be drastically reduced using a biased random sampling from the learned GMM distribution, and the learned models are continually refined and improved as the RRT planning algorithm proceeds. Our proposed method is demonstrated on several example applications and experimental results show marked improvement in computational efficiency over previous approaches.

This paper presents a new approach to the characterization of tactile array sensors that aims to reduce the computational time needed for convergence to obtain a useful estimator for normal and shear forces. This is achieved by breaking up the sensor characterization into two parts: a linear regression portion using multivariate least squares regression, and a nonlinear regression portion using a neural network as a multi-input, multi-output function approximator. This procedure has been termed Least Squares Artificial Neural Network (LSANN). By applying LSANN on the 2nd generation MIT Cheetah footpad, the convergence speed for the estimator of the normal and shear forces is improved by 59.2% compared to using only the neural network alone. The normalized root mean squared error between the two methods are nearly identical at 1.17% in the normal direction, and 8.30% and 10.14% in the shear directions. This approach could have broader implications in greatly reducing the amount of time needed to train a contact force estimator for a large number of tactile sensor arrays (i.e. in robotic hands and skin).

Recent progress in motion planning has made it possible to determine homotopy inequivalent trajectories between an initial and terminal configuration in a robot configuration space. Current approaches have however either assumed the knowledge of differential one-forms related to a skeletonization of the collision space, or have relied on a simplicial representation of the free space. Both of these approaches are currently however not yet practical for higher dimensional configuration spaces. We propose 2D topological task projections (TTPs): mappings from the configuration space to 2-dimensional spaces where simplicial complex filtrations and persistent homology can identify topological properties of the high-dimensional free configuration space. Our approach only requires the availability of collision free samples to identify winding centers that can be used to determine homotopy inequivalent trajectories. We propose the Winding Augmented RRT and RRT* (WA-RRT/RRT*) algorithms using which homotopy inequivalent trajectories can be found. We evaluate our approach in experiments with configuration spaces of planar linkages with 2–10 degrees of freedom. Results indicate that our approach can reliably identify suitable topological task projections and our proposed WA-RRT and WA-RRT* algorithms were able to identify a collection of homotopy inequivalent trajectories in each considered configuration space dimension.

We present hierarchical rejection sampling (HRS) to improve the efficiency of asymptotically optimal sampling-based planners for high-dimensional problems with differential constraints. Pruning nodes and rejecting samples that cannot improve the currently best solution have been shown to improve performance for certain problems. We show that in high-dimensional domains this improvement can be so large that rejecting samples becomes the bottleneck of the algorithm because almost all samples are rejected. This contradicts general wisdom that collision checking is always the bottleneck of sampling-based planners. Only samples in the informed subset of the state space can potentially improve the current solution. For systems without differential constraints the informed subset forms an ellipsoid, which can be parameterized and sampled directly. For systems with differential constraints the informed subset is more complicated and no such direct sampling methods exist. HRS improves the efficiency of finding samples within the informed subset without parameterizing it explicitly. Thus, it can also be applied to systems with differential constraints for which a steering method is available. In our experiments we demonstrate efficiency improvements of an RRT* planner of up to two orders of magnitude.

Learning by demonstration methods have gained considerable interest in human-coupled robot control. It aims at modeling the goal motion trajectories through human demonstration. However, in lower exoskeleton control, the physical human-robot interaction is changing from pilot to pilot or even for one pilot in different walking patterns. This characteristic requires that the exoskeletons should have the ability to learn and adapt the motion trajectories as well as controllers online. This paper presents a novel Hierarchical Interactive Learning (HIL) strategy which reduces the complexity of the exoskeleton sensory system and is able to handle varying interaction dynamics. The proposed HIL strategy is composed of two learning hierarchies, namely, high-level motion learning and low-level controller learning. The Dynamic Movement Primitives (DMPs) combined with Locally Weighted Regression (LWR) are employed to model and learn the motion trajectories, while reinforcement learning (RL) is used to learn the model-based controller. We demonstrate the efficiency of proposed HIL strategy on a single degree-of-freedom (DOF) platform as well as a HUman-powered Augmentation Lower EXoskeleton (HUALEX) system. Experimental results indicate that the proposed HIL strategy is able to handle the varying interaction dynamics with less interaction force between the pilot and the exoskeleton when compared to traditional model-based control algorithms.

In programming by demonstration, generalization is necessary to apply demonstrated motions in novel situations. Many existing trajectory representations have poor generalization capabilities since they are built on trajectory coordinates that depend on the context in which the motion is recorded. In order to generalize, the user is typically required to perform a large number of varied demonstrations. This paper instead emphasizes the usefulness of an invariant trajectory representation to separate essential motion information from context-specific information of the recorded demonstrations. The invariants are interpreted as the control inputs of a dynamical system describing the evolution of the trajectory. New trajectories are generated for novel situations as the solution of a constrained optimal control problem in which context-specific information of the novel situation is encoded in the constraints. Results indicate how, starting from only a single demonstration, new trajectories can be generated in novel situations while maintaining similarity with the original demonstration. Invariance in trajectory representations therefore proves useful to reduce the number of necessary demonstrations to learn and apply new motions.

Motion planning is a fundamental tool in robotics, used to generate collision-free, smooth, trajectories, while satisfying task-dependent constraints. In this paper, we present a novel approach to motion planning using Gaussian processes. In contrast to most existing trajectory optimization algorithms, which rely on a discrete state parameterization in practice, we represent the continuous-time trajectory as a sample from a Gaussian process (GP) generated by a linear time-varying stochastic differential equation. We then provide a gradient-based optimization technique that optimizes continuous-time trajectories with respect to a cost functional. By exploiting GP interpolation, we develop the Gaussian Process Motion Planner (GPMP), that finds optimal trajectories parameterized by a small number of states. We benchmark our algorithm against recent trajectory optimization algorithms by solving 7-DOF robotic arm planning problems in simulation and validate our approach on a real 7-DOF WAM arm.

Robotic manipulation tasks often require the control of forces and torques exerted on external objects. This paper presents a machine learning approach for estimating forces when no force sensors are present on the robot platform. In the training phase, the robot executes the desired manipulation tasks under controlled conditions with systematically varied parameter sets. All internal sensor data, in the presented case from more than 100 sensors, as well as the force exerted by the robot are recorded. Using Transfer Entropy, a statistical model is learned that identifies the subset of sensors relevant for torque estimation in the given task. At runtime, the model is used to accurately estimate the torques exerted during manipulations of the demonstrated kind. The feasibility of the approach is shown in a setting where a robotic manipulator operates a torque wrench to fasten a screw nut. Torque estimates with an accuracy of well below ±1Nm are achieved. A strength of the presented model is that no prior knowledge of the robot's kinematics, mass distribution or sensor instrumentation is required.

We reconsider the problem of reactive navigation in sphere worlds, i.e., the construction of a vector field over a compact, convex Euclidean subset punctured by Euclidean disks, whose flow brings a Euclidean disk robot from all but a zero measure set of initial conditions to a designated point destination, with the guarantee of no collisions along the way. We use power diagrams, generalized Voronoi diagrams with additive weights, to identify the robot's collision free convex neighborhood, and to generate the value of our proposed candidate solution vector field at any free configuration via evaluation of an associated convex optimization problem. We prove that this scheme generates a continuous flow with the specified properties. We also propose its practical extension to the nonholonomically constrained kinematics of the standard differential drive vehicle.

Computationally efficient online learning of non-stationary models remains a difficult challenge. A robust and reliable algorithm could have great impact on problems in learning control. Recent work on combining the worlds of computationally efficient and locally adaptive learning algorithms with robust learning frameworks such as Gaussian process regression has taken a step towards both robust and real-time capable learning systems. However, online learning of model parameters on streaming data - that is strongly correlated, such as data arriving along a trajectory - can still create serious issues for many learning systems. Here we investigate the idea of drifting Gaussian processes which explicitly exploit the fact that data is generated along trajectories. A drifting Gaussian process keeps a history of a constant number of recently observed data points and updates its hyper-parameters at each time step. Instead of optimizing the neighborhood size on which the GP is trained on, we propose to use several - in parallel - drifting GPs whose predictions are combined for query points. We illustrate our approach on synthetically generated data and successfully evaluate on inverse dynamics learning tasks.

Muscle activity monitoring or Electromyography (EMG) is useful in gait analysis, injury prevention, computer or robot interfaces and assisting patients with communication difficulties. However, EMG is typically invasive or obtrusive, expensive and difficult to use for untrained users. A possible solution is textile-based surface EMG (sEMG) integrated into clothing and used as a wearable device. This is, however, challenging due to (i) uncertainties in the electrical properties of conductive threads used to construct electrodes, (ii) imprecise fabrication technologies (e.g., embroidery, sewing), and (iii) a lack of standardisation in the choice of design variables. This paper, for the first time, provides a design guide for such sensors by performing a thorough examination of the effect of design variables on sEMG quality. Electrical characterisation and sEMG measurements are performed, considering the effects of manufacturing imprecision. Results show that the imprecisions in digital embroidery lead to a trade-off between low electrode resistance and high consistency. An optimum set of variables for this trade-off is identified and tested with sEMG during a variable force isometric grip exercise with n=6 participants and compared with traditional gel-based electrodes. Results show that thread-based electrodes provide a similar level of sensitivity to force variation as gel-based electrodes with about 90% correlation with expected linear behaviour.

We investigate the cooperative pursuit of an evader by a group of quadrotors in an environment with no-fly zones. While the pursuers cannot enter into no-fly zones, the evader may freely move through zones to avoid capture. Once the evader enters a no-fly zone, the pursuers calculate a reachable set of evader positions. Using tools from Voronoi-based coverage control applied to the evader's reachable set, we provide an algorithm that distributes the pursuers around the zone's boundary and minimizes the capture time once the evader leaves the no-fly zone. Robust model predictive control (RMPC) tools are used to control the quadrotors and to ensure that they always remain in free space. We demonstrate the performance of our proposed algorithms through a series of experiments on KMEL Nano+ quadrotors.

Whole body tactile perception via tactile skins offers large benefits for robots in unstructured environments. To fully realize this benefit, tactile systems must support real-time data acquisition over a massive number of tactile sensor elements. We present a novel approach for scalable tactile data acquisition using compressed sensing. We first demonstrate that the tactile data is amenable to compressed sensing techniques. We then develop a solution for fast data sampling, compression, and reconstruction that is suited for tactile system hardware and has potential for reducing the wiring complexity. Finally, we evaluate the performance of our technique on simulated tactile sensor networks. Our evaluations show that compressed sensing, with a compression ratio of 3 to 1, can achieve higher signal acquisition accuracy than full data acquisition of noisy sensor data.

We present an automatic method for interactive control of physical humanoid robots based on high-level tasks that does not require manual specification of motion trajectories or specially-designed control policies. The method is based on the combination of a model-based policy that is trained off-line in simulation and sends high-level commands to a model-free controller that executes these commands on the physical robot. This low-level controller simultaneously learns and adapts a local model of dynamics on-line and computes optimal controls under the learned model. The high-level policy is trained using a combination of trajectory optimization and neural network learning, while considering physical limitations such as limited sensors and communication delays. The entire system runs in real-time on the robot's computer and uses only on-board sensors. We demonstrate successful policy execution on a range of tasks such as leaning, hand reaching, and robust balancing behaviors atop a tilting base on the physical robot and in simulation.

This paper presents a new approach to C-space exploration and path planning for robotic manipulators using the structure named bur of free C-space. This structure builds upon the so-called bubble, which is a local volume of free C-space, easily computed using the distance information in the workspace. We show how the same distance information can be used to compute the bur that can reach substantially beyond the boundary of the bubble. It is shown how burs can be used to form a rapidly exploring bur tree (RBT): a space-filling tree that resembles RRT. Such a structure can easily be used within a suitably tailored path planning algorithm. Simulation study shows how the RBT-based algorithm outperforms the classical RRT-based method.