IROS 2017 Papers

Non-prehensile manipulation, such as pushing, is an important function for robots to move objects and is sometimes preferred as an alternative to grasping. However, due to unknown frictional forces, pushing has been proven a difficult task for robots. We explore the use of reinforcement learning to train a robot to robustly push an object. In order to deal with the sample complexity of training such a method, we train the pushing policy in simulation and then transfer this policy to the real world. In order to ease the transfer from simulation, we propose to use modularity to separate the learned policy from the raw inputs and outputs; rather than training “end-to-end,” we decompose our system into modules and train only a subset of these modules in simulation. We further demonstrate that we can incorporate prior knowledge about the task into the state space and the reward function to speed up convergence. Finally, we introduce “reward guiding” to modify the reward function and further reduce the training time. We demonstrate, in both simulation and real-world experiments, that such an approach can be used to reliably push an object from many initial positions and orientations. Videos available at https://goo.gl/B7LtY3.

Planning safe, high-speed trajectories in unknown environments remains a major roadblock on the way toward achieving fast autonomous flight. Current state-of-the-art planning approaches use sampling-based methods or trajectory optimization to obtain fast trajectories, whose safety is evaluated by taking into account the current state estimate of the environment. In unknown environments, however, this leads to numerous stops caused by the need for re-planning the trajectory due to unexpected obstacles. In this paper, we propose to use an active perception paradigm for planning. We predict the future uncertainty of the map and optimize trajectories to minimize re-planning risk. This leads to faster and safer trajectories. We evaluate the proposed planning approach in a series of simulation experiments, which show that we are able to achieve safer trajectories with a smaller number of re-planning stops and faster speeds.

This paper describes an approach that enables the four-limbed robot “WAREC-1” to climb up and down vertical ladders stably. First, the four-limbed robot “WAREC-1” is introduced and dynamic stability conditions in multi-mass model for ladder climbing are proposed as the basis of judging whether a four-limbed robot is stable or not while climbing a vertical ladder. According to the proposed stability conditions, 3 different types of moment will directly affect the stability of the robot on a ladder: gravitational moment, inertial moment and reaction force moment. With the analysis of these 3 kinds of moments and the relationship among them, stability control methods are proposed to maintain stability of the robot on a ladder to the greatest degree and avoid their mutual interference. Combining with the stability conditions and stability control proposed, stable motion planning of climbing up and down a vertical ladder, a motion planning method proposed by the authors that allows independent path and time planning in trajectory planning is also applied to reinforce the efficiency of the stability control. Eventually, results from the simulation and physical robot verify the validity of the proposed control methods.

Planetary exploration involves frequent scientific reformulation and replanning. It is limited by communication constraints and to overcome this limitation, this paper formulates the process as a collaboration in which the human scientist and the robot work together to fill in gaps in knowledge to make discoveries. It introduces the science hypothesis map as the probabilistic structure in which scientists initially describe their abstract beliefs and hypotheses, and in which the state of this belief evolves as the robot makes raw measurements. It discusses how to incorporate path planning for maximizing scientific information gain, which is efficiently computed. As proof of concept, this paper describes a geologic exploration problem where a robot uses a spectrometer to infer the geologic composition of different regions in a mining district at Cuprite, Nevada. It shows that the science hypothesis map can infer geologic units with high accuracy, and that exploration using information gain-based path planning has better performance than exploration with conventional science-blind algorithms.

The main contribution of this paper is a planar scan matching algorithm that makes use of the incident angle of a scan point as a feature to enhance the robustness to large relative transformations, particularly in orientation. A new definition of the incident angle is introduced and its consistency with respect to relative transformation between scans is demonstrated. A method that uses the Radial distance and Incident Angle (RIA) for scan alignment, that can efficiently estimate the relative transformation between two scans, is proposed. This approach is designed as a preprocessor for fine scan matching and it is able to pre-align scans that have large angular but limited translational displacement. Scans collected with aggressive robot motion in two typical indoor environments as well as benchmark datasets, are used to demonstrate the effectiveness of RIA when compared with a range of scan matching algorithms reported in the literature.

Biologically inspired by the hopping performance of kangaroo, this paper extends the traditional Spring-loaded Inverted Pendulum (SLIP) model by adding an actuation at hip to composite a tail-actuated SLIP (TSLIP) model as an abstract template for gait controller design. To deal with the intrinsic nonlinearity of stance dynamics in the TSLIP model, an analytical approximation is derived by virtue of perturbation technique with gravity correction. Employing the derived solution to construct the apex return map, a gait controller is further devised with a two-layer nonlinear optimization scheme. The outer loop optimizes the tail motion during stance by matching the energy variation between the current and target apex state while the inner loop subsequently selects the optimum touchdown angle by minimizing the difference between the predictive and target apex vector from stride to stride. Additionally, an extended control strategy that embodies a time-dependent pre-positioned policy for swing-leg working in conjunction with the active tail is devised, requiring no priori knowledge of the ground truth to enhance to hopping performance of the TSLIP system. The simulation results have demonstrated the effectiveness of the proposed control strategy for tailed hopping system.

Place recognition based on landmarks or features is an important problem occurring in localization, mapping, computer vision and point cloud processing. In this paper, we present GLAROT-3D, a translation and rotation invariant 3D signature based on geometric relations. The proposed method encodes into a histogram the pairwise relative positions of keypoint features extracted from 3D sensor data. Since it relies only on geometric properties and not on specific feature descriptors, it does not require any prior training or vocabulary construction and enables lightweight comparisons between landmark maps. The similarity of two point maps is computed as the distance between the corresponding rotated histograms to achieve rotation invariance. Histogram rotation is enabled by efficient orientation histogram based on sphere cubical projection. The performance of GLAROT has been assessed through experiments with standard benchmark datasets.

Robots have the potential to provide tremendous support to disabled and elderly people in their everyday tasks, such as dressing. Many recent studies on robotic dressing assistance usually view dressing as a trajectory planning problem. However, the user movements during the dressing process are rarely taken into account, which often leads to the failures of the planned trajectory and may put the user at risk. The main difficulty of taking user movements into account is caused by severe occlusions created by the robot, the user, and the clothes during the dressing process, which prevent vision sensors from accurately detecting the postures of the user in real time. In this paper, we address this problem by introducing an approach that allows the robot to automatically adapt its motion according to the force applied on the robot's gripper caused by user movements. There are two main contributions introduced in this paper: 1) the use of a hierarchical multi-task control strategy to automatically adapt the robot motion and minimize the force applied between the user and the robot caused by user movements; 2) the online update of the dressing trajectory based on the user movement limitations modeled with the Gaussian Process Latent Variable Model in a latent space, and the density information extracted from such latent space. The combination of these two contributions leads to a personalized dressing assistance that can cope with unpredicted user movements during the dressing while constantly minimizing the force that the robot may apply on the user. The experimental results demonstrate that the proposed method allows the Baxter humanoid robot to provide personalized dressing assistance for human users with simulated upper-body impairments.

Differential drive robots, such as robotic vacuums, often have at least two motion primitives: the ability to travel forward in straight lines, and rotate in place upon encountering a boundary. They are often equipped with simple sensors such as contact sensors or range finders, which allow them to measure and control their heading angle with respect to environment boundaries. We aim to find minimal control schemes for creating stable, periodic “patrolling” dynamics for robots that drive in straight lines and “bounce” off boundaries at controllable angles. As a first step toward analyzing high-level mobile robot dynamics in more general environments, we analyze the location and stability of periodic orbits in regular polygons. The contributions of this paper are: 1) proving the existence of periodic trajectories in n-sided regular polygons and showing the range of bounce angles that will produce such trajectories; 2) an analysis of their stability and robustness to modeling errors; and 3) a closed form solution for the points where the robot collides with the environment boundary while patrolling. We present simulations confirming our theoretical results.

We revisit the classical problem of estimating the dynamic parameters of an unknown payload rigidly held by the robot end effector. The approach relies on the analysis of the symbolic expressions of the robot dynamic coefficients (i.e., combinations of dynamic parameters) when working with and without payload, with no special assumptions on payload structure. The linearity of the associated changes in the dynamic coefficients due to the payload addition is exploited so as to estimate the gravity and inertial parameters of the payload. The procedure is illustrated in simulation on a planar 2R robot with asymmetric payload and through experiments on a 7R KUKA LWR arm with medium payloads. Accurate estimates of non-inertial payload parameters can be obtained even by running the identification scheme on few small motions in a restricted area. The results are shown to be useful for improving the sensorless collision detection capabilities of a robot arm in the presence of an a priori unknown payload.

This paper introduces two well-accepted path-following control methods, namely Vector Field Method (VF) and Line-Of-Sight Method (LOS) for Unmanned Surface Vehicles (USVs). We provide a comprehensive study of each algorithm, which includes investigating their mathematical origins, performing simulation evaluations and carrying out real-world field tests in adverse weather conditions. We compare different characteristics of the two methods, and the successful field trial results demonstrate their accuracy and robustness considering unexpected environmental influences. The presented work can assist USV practitioners to decide on a proper strategy for completion of a specific USV mission.

Microrobots have great potential for microassembly and non-invasive surgery applications. Motivated by studies proposing MRI-guided drug delivery to tumor cells using magnetic micro carriers, this paper studies two major challenges of this problem: (i) microrobot swarm trajectory generation, and (ii) swarm aggregation using a global input. We propose an augmented RRT for trajectory generation to reduce environment interference, and a divide-and-conquer algorithm for swarm aggregation to improve performance. Simulations demonstrate the utility of these approaches in comparison to alternate heuristics. Our trajectory generation and aggregation strategies are implemented on a swarm of ferromagnetic microparticles in oil using a 6-coil electromagnetic system with image feedback.

Robotic walkers are a particular class of devices used to assist users with physical or cognitive impairments in their navigation of large public spaces. In this context, a guidance mechanism is a controller that steers the user towards the desired path when he/she deviates. Passive guidance mechanisms do not directly propel the vehicle and leave the user in control of his/her walk. The inherent safety of passive guidance and the intuitive behaviour of the device make this class of solutions preferable to any other for robot assisted walking. The possible ways to obtain a passive behaviour in robotic walkers either require complex and expensive sensors or generate potentially rough (bang-bang) manoeuvres that are detrimental to the user's comfort. The contribution of this work is a path following controller that simulates passivity using a pair of active motors operating on the rear wheels of the walker. The system estimates and tracks the velocity desired by the user and the motors generate a rotational torque only when a turn toward the path is required. The proposed solution delivers high levels of comfort, does not rely on expensive hardware and preserves all the important properties of a passive mechanism.

We develop a novel passivity-based control framework of a manipulator-stage system sitting on a vertical flexible beam. We first model the vertical flexible beam by using Euler-Bernoulli theory and modal approximation while utilizing certain boundary conditions to take into account the effect of the gravity and interaction wrench between the manipulator-stage system and the beam. To separately achieve two control objectives, i.e. the joint tracking and the vibration suppression, we utilize the passive decomposition to split the dynamics into: 1) the stage and beam dynamics; and 2) its orthogonal complement, which converges to the robotic manipulator dynamics as the vibration is subdued. We further show that the linearized stage-beam dynamics is controllable and design an LQR control to stabilize the beam vibration only by using the stage motion. We also design control for the orthogonal complement dynamics in such a way that it can attain the manipulator joint tracking asymptotically as the vibration is suppressed. Experiments are also performed to verify the manipulator joint tracking while the vibration is suppressed internally via the stage motion.

In this paper, we introduce a concept of passive knee exoskeleton for cycling assistance. Considering a knee moment and a knee angle varying with a pedal crank angle, the knee extension moment can be supported by a torsion spring storing energy from knee flexion in order to release it as the knee is extended. The reduction of knee extension effort is corresponding to the torsion spring stiffness and activation range. Exoskeleton prototypes were developed for the concept validation. A crossing four-bar mechanism was chosen for the knee joint to provide kinematic compatibility covering extreme knee flexion. Constant power cycling experiment was performed on a cycling trainer by a healthy subject wearing the exoskeletons on both legs. With the torsion spring support, the surface electromyography recorded from some major leg muscles shows the decrease of knee extensor muscle activity as the leg is moving around the pedal crank top dead center. Verifying the reduction of leg muscle fatigue over repetitive contractions in multiple subjects is our future plan of study.

This paper proposes a new method for human posture recognition from top-view depth maps on small training datasets. There are two strategies developed to leverage the capability of convolution neural network (CNN) in mining the fundamental and generic features for recognition. First, the early layers of CNN should serve the function to extract feature without specific representation. By applying the concept of transfer learning, the first few layers from the pre-learned VGG model can be used directly without further fine-tuning. To alleviate the computational loading and to increase the accuracy of our partially transferred model, a cross-layer inheriting feature fusion (CLIFF) is proposed by using the information from the early layer in fully connected layer without further processing. The experimental result shows that combination of partial transferred model and CLIFF can provide better performance than VGG16 [1] model with re-trained FC layer and other hand-crafted features like RBPs [2].

We present PRVO, a probabilistic variant of Reciprocal Velocity Obstacle (RVO) for decentralized multi-robot navigation under uncertainty. PRVO characterizes the space of velocities that would allow each robot to fulfill its share in collision avoidance with a specified probability. PRVO is modeled as chance constraints over the velocity level constraints defined by RVO and takes into account the uncertainty associated with both state estimation as well as the actuation of each robot. Since chance constraints are in general computationally intractable, we propose a series of reformulations which when combined with time scaling based concepts leads to a closed form characterization of solution space of PRVO for a given probability of collision avoidance. We validate our formulation through numerical simulations in which we highlight the advantages of PRVO over the related existing formulations.

A new navigation system for outdoor person following at higher speeds (maximum speed ≈2.5 m/s) is proposed. A combination of global and local path planning, and path following control, allows a robot to follow a jogger in various outdoor scenarios, including highly dynamical environments with pedestrians. Our fast path planning provides smooth and obstacle-free paths, even at higher speeds, in the presence of dynamic obstacles and in narrow passages. Our control algorithm allows precise following of the planned paths on different terrain types, including snow and ice. The mobile robot used in the experiments is the skid-steered robot Robotnik Summit XL.

In this paper we address the problem of assigning a set of tasks to a set of Automated Guided Vehicles (AGVs), in a conflict-free manner. Specifically, we consider a system of multiple AGVs, moving along a predefined roadmap, and utilized for transportation of goods in automated warehouses. Sequential application of task assignment and path planning often gives rise to pathological situations, such as deadlocks, in which AGVs block each other, thus preventing tasks completion. In this paper we propose a method for assigning tasks while taking into account the subsequent path planning, encoding possible conflicts into a conflict graph, that is subsequently utilized for defining constraints of an optimization problem. Simulations are performed on maps of real industrial environments, to compare the proposed method with traditional task assignment.