ICRA 2017 论文列表

In this paper, we present a method to improve the navigation of tethered underwater vehicles by computing optimal paths that prevent their tethers from becoming entangled in obstacles. To accomplish this, we define the Non-Entangling Travelling Salesperson Problem (NE-TSP) as an extension of the Travelling Salesperson Problem with a non-entangling constraint. We compute the optimal solution to the NE-TSP by constructing a Mixed Integer Programming model, leveraging homotopy augmented graphs to plan an optimal trajectory through a set of inspection points, while maintaining a non-entangling guarantee. To avoid the computational expense of computing an optimal solution to the NE-TSP, we also introduce several methods to compute near-optimal solutions. In a set of simulated trials, our method was able to plan optimal non-entangling paths through a variety of environments. These results were then validated in a set of pool and field trials using a Seabotix vLBV300 underwater vehicle. The paths generated by our method were then compared to human-generated paths.

In this paper we propose a method for motion planning and feedback control of hybrid, dynamic, and non-prehensile manipulation tasks. We outline five subproblems to address this: determining a set of manipulation primitives, choosing a sequence of tasks, picking transition states, motion planning for each individual primitive, and stabilizing each mode using feedback control. We apply the framework to plan a sequence of motions for manipulating a block with a planar 3R manipulator. We demonstrate preliminary experimental results for a block resting on the manipulator with a desired goal state on a ledge outside of the robot's workspace. The planned primitives reorient the block using a series of fixed, rolling, and sliding contact modes, and throw it to the goal state.

Intelligent robots and machines are becoming pervasive in human populated environments. A desirable capability of these agents is to respond to goal-oriented commands by autonomously constructing task plans. However, such autonomy can add significant cognitive load and potentially introduce safety risks to humans when agents behave in unexpected ways. Hence, for such agents to be helpful, one important requirement is for them to synthesize plans that can be easily understood by humans. While there exists previous work that studied socially acceptable robots that interact with humans in “natural ways”, and work that investigated legible motion planning, there is no general solution for high level task planning. To address this issue, we introduce the notions of plan explicability and predictability. To compute these measures, first, we postulate that humans understand agent plans by associating abstract tasks with agent actions, which can be considered as a labeling process. We learn the labeling scheme of humans for agent plans from training examples using conditional random fields (CRFs). Then, we use the learned model to label a new plan to compute its explicability and predictability. These measures can be used by agents to proactively choose or directly synthesize plans that are more explicable and predictable to humans. We provide evaluations on a synthetic domain and with a physical robot to demonstrate the effectiveness of our approach.

The goal of Piccolissimo is to create a small, simple, and self-powered flying vehicle. Piccolissimo has just one motor and two rigid bodies, which are propellers that spin in opposite directions. The mass distribution and relative rotor speeds are designed to maintain passive stability in hover. Cartesian velocity control is obtained by introducing an asymmetry in the rotation axis of the two rotating bodies and pulsing the thrust at appropriate times. The dynamic model for this device is developed and presented, highlighting the terms that enable a workable design. Two devices are discussed: a vertically controllable version that is 28 mm in the largest dimension and Cartesion controllable version that is 39 mm in the largest dimension.

Path planning for mobile robots requires rapidly finding collision-free trajectories in an uncertain and changing environment. Full collision checking with detailed, online-revised representations of the robot and world imposes a delay that undermines reactive obstacle avoidance. As a result, reactive vision-based approaches make various assumptions to arrive at simplified representations, such as circular or spherical robot shapes reducible to point masses, or obstacles that always rise from the ground. We seek to avoid these problems by modeling the robot directly in perception space so that collisionfree trajectories can be sought in a consistent representation with minimal processing needs. Here perception space refers to the depth space image measurements available by modern consumer range sensors. We hallucinate a robot navigating through the world and synthesize depth images of its path for comparison against the directly sensed depth images of the local world. The approach performs collision checking in a 3D volume but only requires 2D image comparisons. Experiments show that an implementation is able to negotiate an obstacle course consisting of miscellaneous objects in real-time.

Many interesting natural phenomena are sparsely distributed and discrete. Locating the hotspots of such sparsely distributed phenomena is often difficult because their density gradient is likely to be very noisy. We present a novel approach to this search problem, where we model the co-occurrence relations between a robot's observations with a Bayesian nonparametric topic model. This approach makes it possible to produce a robust estimate of the spatial distribution of the target, even in the absence of direct target observations. We apply the proposed approach to the problem of finding the spatial locations of the hotspots of a specific phytoplankton taxon in the ocean. We use classified image data from Imaging FlowCytobot (IFCB), which automatically measures individual microscopic cells and colonies of cells. Given these individual taxon-specific observations, we learn a phytoplankton community model that characterizes the co-occurrence relations between taxa. We present experiments with simulated robot missions drawn from real observation data collected during a research cruise traversing the US Atlantic coast. Our results show that the proposed approach outperforms nearest neighbor and k-means based methods for predicting the spatial distribution of hotspots from in-situ observations.

It is natural for humans to work with abstract plans which are often an intuitive and concise way to represent a task. However, high level task descriptions contain symbols and concepts which need to be grounded within the environment if the plan is to be executed by an autonomous robot. The problem of learning the mapping between abstract plan symbols and their physical instances in the environment is known as the problem of physical symbol grounding. In this paper, we propose a framework for Grounding and Learning Instances through Demonstration and Eye tracking (GLIDE). We associate traces of task demonstration to a sequence of fixations which we call fixation programs and exploit their properties in order to perform physical symbol grounding. We formulate the problem as a probabilistic generative model and present an algorithm for computationally feasible inference over the proposed model. A key aspect of our work is that we estimate fixation locations within the environment which enables the appearance of symbol instances to be learnt. Instance learning is a crucial ability when the robot does not have any knowledge about the model or the appearance of the symbols referred to in the plan instructions. We have conducted human experiments and demonstrate that GLIDE successfully grounds plan symbols and learns the appearance of their instances, thus enabling robots to autonomously execute tasks in initially unknown environments.

In this paper we present a novel direct visual-inertial odometry algorithm, for estimating motion in unknown environments. The algorithm utilizes image patches extracted around image features, and formulates measurement residuals in the image intensity space directly. One key characteristic of the proposed method is that it models the true irradiance at each pixel as a random variable to be estimated and marginalized out. The formulation of the photometric residual explicitly accounts for the camera response function and lens vignetting (which can be calibrated in advance), as well as unknown illumination gains and biases, which are estimated on a per-feature or per-image basis. We present a detailed evaluation of our algorithm on 50 datasets with high-precision ground truth, which amount to approximately 1.5 hours of localization data. Through a direct comparison with a point-feature based method, we demonstrate that the use of photometric residuals results in increased pose estimation accuracy, with approximately 23% lower estimation errors, on average.

We present an off-board phase estimator and controller for leg position near the resonance of the Harvard Ambulatory MicroRobot's (HAMR) two degree-of-freedom transmission. This control system is a first step towards leveraging the significant increase in stride length at transmission resonance for faster and more efficient locomotion. We experimentally characterize HAMR's transmission and determine that actuator phase is a sufficient proxy for leg phase across the range of useful operating frequencies (1-120Hz). An estimator is developed to determine actuator phase using off-board position sensors and it converges within a cycle on average. We also fit a nonlinear dynamic model of the transmission to the experimental data, and utilize the model to determine a suitable open-loop resonant leg trajectory and define feed forward control inputs. This resonant (100Hz) trajectory is theoretically 50% more efficient than pre-resonant high speed running trajectories. The controller converges to this trajectory in 0.05 ± 0.02 seconds (5.3 ± 2.4 cycles) in air, and in 0.05 ± 0.01 seconds (4.7 ± 0.6 cycles) under perturbations that approximate ground contact.

We consider the problem of monitoring stochastic, time-varying events occurring at discrete locations. Our problem formulation extends prior work in persistent surveillance by considering the objective of maximizing event detections in unknown, dynamic environments where the rates of events are time-inhomogeneous and may be subject to abrupt changes. We propose a novel monitoring algorithm that effectively strikes a balance between exploration and exploitation as well as a balance between remembering and discarding information to handle temporal variations in unknown environments. We present an analysis proving the long-run average optimality of the policies generated by our algorithm under the assumption that the total temporal variations are sub-linear. We present simulation results demonstrating the effectiveness of our algorithm in several monitoring scenarios inspired by real-world applications, and its robustness to both continuous-random and abrupt changes in the statistics of the observed processes.

We consider a visibility-based pursuit-evasion problem in which a single robot with an omnidirectional but unreliable sensor moving through an environment must systematically search that environment to detect an unpredictably moving target. A common assumption in visibility-based pursuit-evasion is that the sensors used to detect the evader are perfectly reliable. That is, any evader that moves within view of the pursuer for any interval of time will be detected. This assumption is problematic because, when implemented on real sensor systems, such plans cannot account for the possibility of short-term false negative errors in evader detection. This paper addresses this limitation by introducing a model based on the idea of pessimal unoccluded distance to reason about the degree of plausibility that the evader may be concealed within each occluded region. We describe a decomposition of the environment that fully characterizes the opportune moment for an evader to take advantage of sensor error. Furthermore, we present a complete algorithm that solves the active problem of planning a search for a pursuer which maximizes the distance that the evader must travel through the pursuer robot's sensor footprint.

Adaptive cable-driven parallel robots can adjust the position of one or more pulley blocks to optimize performance within a given workspace. Because of their augmented kinematic redundancy, adaptive systems have several advantages over their traditional counterparts featuring the same numbers of cables. In this paper, we explore the application of adaptive cable-driven robots to cable-suspended camera systems. Performance of the traditional and of the adaptive designs are analyzed, using dexterity and stiffness as performance metrics. Results show superior performance of the adaptive design compared to the traditional system. An illustrative design problem for adaptive cable-suspended camera systems is also presented and solved.

A wheeled inverted pendulum cannot balance without control, and it is important to save motion in a control range. In some situations, however, to avoid a collision, the maximum acceleration and deceleration must be applied. A motion-planning method for a wheeled inverted pendulum (“body” hereafter)-which uses full performance from all attitudes under hardware constraints-was developed. The developed method involves two steps. First, a linear combination of zeroth to fourth differentials of parameters based on the motion equation of a wheeled inverted pendulum is obtained. Second, the arbitrary attitude of the wheeled inverted pendulum and hardware constraints are expressed on a phase plane containing angular rate of body motion and an angular inclination of the body. By executing these two steps, a motion plan can be expressed on the phase plane. A simulation based on a model confirmed that the movement plan formulated by this technique can be followed with no error.

We present PennCOSYVIO, a new challenging Visual Inertial Odometry (VIO) benchmark with synchronized data from a VI-sensor (stereo camera and IMU), two Project Tango hand-held devices, and three GoPro Hero 4 cameras. Recorded at UPenn's Singh center, the 150m long path of the hand-held rig crosses from outdoors to indoors and includes rapid rotations, thereby testing the abilities of VIO and Simultaneous Localization and Mapping (SLAM) algorithms to handle changes in lighting, different textures, repetitive structures, and large glass surfaces. All sensors are synchronized and intrinsically and extrinsically calibrated. We demonstrate the accuracy with which ground-truth poses can be obtained via optic localization off of fiducial markers. The data set can be found at https://daniilidis-group.github.io/penncosyvio/.

We present a policy search method for learning complex feedback control policies that map from high-dimensional sensory inputs to motor torques, for manipulation tasks with discontinuous contact dynamics. We build on a prior technique called guided policy search (GPS), which iteratively optimizes a set of local policies for specific instances of a task, and uses these to train a complex, high-dimensional global policy that generalizes across task instances. We extend GPS in the following ways: (1) we propose the use of a model-free local optimizer based on path integral stochastic optimal control (PI2), which enables us to learn local policies for tasks with highly discontinuous contact dynamics; and (2) we enable GPS to train on a new set of task instances in every iteration by using on-policy sampling: this increases the diversity of the instances that the policy is trained on, and is crucial for achieving good generalization. We show that these contributions enable us to learn deep neural network policies that can directly perform torque control from visual input. We validate the method on a challenging door opening task and a pick-and-place task, and we demonstrate that our approach substantially outperforms the prior LQR-based local policy optimizer on these tasks. Furthermore, we show that on-policy sampling significantly increases the generalization ability of these policies.

This paper addresses the problem of path following for a biomimetic underwater vehicle (BUV) propelled by undulatory fins with uncertain model and unknown disturbance. The mechanical structure of the BUV is briefly described. Moreover, the general kinematics and dynamics models of the vehicle are presented and the path following problem is formulated. The controller combining line-of-sight (LOS) guidance system with active disturbance rejection control (ADRC) technique is designed to maneuver the BUV to follow a predefined parameterized curve. Specifically, a guidance system based on LOS principle is implemented to decouple the multi-variable system to steer the surge speed and the course respectively. Furthermore, in order to deal with model uncertainty, ADRC is used in development of the surge speed controller and the course controller. Finally, simulations and experimental results validated the performance of the proposed path following control scheme.

Pseudo-omnidirectional robots with individually steerable wheels ofïer a good balance between mobility, robustness and load-carrying capacity. However, accurate synchronization of the wheels' steering and rolling speeds is necessary to prevent energy-loss, mechanical stress and wheel slippage due to actuator infighting. This has proven to be a complex problem, especially when the wheels are not fixed to the robot body but are instead connected to it via actuated chains. This paper presents a mathematically simple method consisting of closed-form equations for path following and synchronizing the steering and rolling speeds of planar robots with a variable or fixed footprint, while respecting the velocity limits of each wheel's steering and rolling actuators. The presented method is thoroughly explained and simulation results are presented to show its performance.

A new nonlinear control law for path following with skid-steered mobile robots is proposed. A terrain dependent kinematic model is utilized in path coordinates, and the kinematic parameters are experimentally evaluated. A kinematic path following control is developed using the Lyapunov approach. A separate linear velocity control is then proposed, taking reachable curvatures and actuator saturation into account. The proposed approach is experimentally evaluated in different terrain scenarios, and compared with two other state-of-the-art algorithms. The skid-steered vehicle used for the experiments is the Robotnik Summit XL, a well known commercial mobile robot.

Direct force control of robots is challenging, particularly since the interaction with the environment can render the robot unstable. This paper presents the results of novel approaches for passivity-based stability for a particular direct force control method, namely explicit force control. A step-by-step procedure to passivate and stabilise the control loop is presented and it explains how Time Domain Passivity Approach, a passivity-based tool widely used in teleoperation and haptics has been extended and applied in explicit force control. The electrical circuit and network-port representations derived in the process allows the analytical evaluation of the system and can be applied in other control architectures as well. The stability methods are presented both qualitatively and quantitatively with simulations and hardware experiments. A discussion about the results obtained and the energy behavior is also provided. Results are promising and suggest that these methods can be used for stable and high-bandwidth force control of robotic manipulators.