IROS 2016 Papers

Multirotor helicopters are expected to be utilized various tasks including rescue missions and surveillance. For those missions, sensors are equipped with helicopters in order to recognize the environment, and auditory information is one of such information that can be utilized to find the target sound source even if it is occluded by objects. One of the difficulty comes from the fact that the noise generated by rotating rotors distorts the target signal significantly, which leads inaccurate estimate of the target source direction. Besides, the estimate of the attitude and the position of the helicopter is inaccurate, which is another technical issue. This paper proposes a robust sound source position estimation taking such uncertainty into account, and experiment with real flight tests showed that the helicopter was able to estimate the target position within about 2m accuracy.

In recent years, bowden-cable transmissions have been developed and utilized widely in many robotic applications due to advantages in durability, lightweight, safety, and flexibility. Especially, over the last decade, a substantial number of soft wearable exoskeletons using bowden cables for motion transmission have been designed for human assistance, empowerment and rehabilitation. The major advantage of soft assistive devices driven by bowden-cable transmissions is to allow decentralizing the actuation stages proximally such that their mass has the least effect on the end-effector. Besides the advantage, the main drawback of the bowden cable-driven system comes from the presence of nonlinearities such as friction and backlash hysteresis that affects their control accuracy. Hence, in this paper, we introduce a mathematical model for backlash hysteresis and propose a solution based on the nonlinear adaptive control to compensate for the backlash effect. The backlash hysteresis model and control scheme are validated first on a custom-designed test bench and then applied to control a soft exoskeleton in a preliminary human trial.

In this paper, we propose a hierarchical pose graph optimization algorithm, which hierarchically divides a large pose graph into subgraphs and solves the optimization problem for each subgraph independently. Applying a modified graph partitioning algorithm, normalized cut, the original graph could be partitioned into subgraphs, which are conditionally independent on a set of key nodes. A modified normalized cut algorithm is applied to automatically partition a pose graph into several subgraphs, which are independent on each other when conditioned on a small number of keynodes. Preserving keynodes only, a simplified upper level graph is generated by a pose graph sparsification algorithm. Given the optimization results of all keynodes, which are obtained by solving the upper level graph, each subgraph in the lower level can be solved efficiently without concerning other subgraphs. The scale of each optimization is limited during graph partitioning. Therefore, the efficiency of the optimization is improved. Experiments with both public standard datasets and a dataset collected by an autonomous guided vehicle (AGV) system are conducted to test our algorithm.

Automated driving relies on fast, recent, accurate, and highly available pose estimates. A single localization system, however, can commonly ensure this only to some extent. In this paper, we propose a multi-sensor fusion approach that resolves this by combining multiple localization systems in a plug and play manner. We formulate our approach as a sliding window pose graph and enforce a particular graph structure which enables efficient optimization and a novel form of marginalization. Our pose fusion approach scales from a filtering-based to a batch solution by increasing the size of the sliding window. We evaluate our approach on simulated data as well as on real data gathered with a prototype vehicle and demonstrate that our solution runs comfortably at 20 Hz, provides timely estimates, is accurate, and yields a high availability.

We propose an approach for emptying of bin using a combination of Image and Range sensor. Offering a complete solution: calibration, segmentation and pose estimation, along with approachability analysis for the estimated pose. The work is novel in the sense that the objects to be picked are featureless and uniformly black in colour, hence existing approaches are not directly applicable. A key point involves optimal utilization of range data acquired from the laser scanner for 3-D segmentation using localized geometric information. This information guides segmentation of the image for better object pose estimation, used for pick-and-drop. We analytically assure the approachability of the object to avoid collision of the manipulator with the bin. Disturbance of objects caused during pick up has been modelled, which allows pickup of multiple pellets based on information from a single range scan. This eliminates the necessity of repeated scanning and data conditioning. The proposed method offers high object detection rate and pose estimation accuracy. The innovative techniques aimed at reducing the average pickup time makes it suitable for robust industrial operation.

We present a drift-free pose estimation scheme for rigid body and its supporting platform by fusing only two gyroscopes and the relative complementary measurements. The fusion design not only provides robust relative attitude estimation between the rigid body and the platform, but also is capable of identifying partial global absolute attitudes without capturing any absolute attitude information. The pose estimation is built on a special design of the coupled kinematic model with the relative measurements between the rigid body and its supporting platform. We compare the fusion design with an alternative kinematic model and the posterior Cramer-Rao bound analyses are presented to show the completely different estimation performances. An extended Kalman filter (EKF) implementation of the fusion design is presented for the bicycle riding application.

Existing simultaneous localization and mapping (SLAM) algorithms are not robust in challenging low-texture environments because there are only few salient features. The resulting sparse or semi-dense map also conveys little information for motion planning. Though some work utilize plane or scene layout for dense map regularization, they require decent state estimation from other sources. In this paper, we propose real-time monocular plane SLAM to demonstrate that scene understanding could improve both state estimation and dense mapping especially in low-texture environments. The plane measurements come from a pop-up 3D plane model applied to each single image. We also combine planes with point based SLAM to improve robustness. On a public TUM dataset, our algorithm generates a dense semantic 3D model with pixel depth error of 6.2 cm while existing SLAM algorithms fail. On a 60 m long dataset with loops, our method creates a much better 3D model with state estimation error of 0.67%.

Localization is a key capability for autonomous vehicles especially in urban scenarios. We propose the use of pole-like landmarks as primary features in these environments, as they are distinct, long-term stable and can be detected reliably with a stereo camera system. Furthermore, the resulting map representation is memory efficient, allowing for easy storage and on-line updates. The localization is performed in real-time by a stereo camera system as a main sensor, using vehicle odometry and an off-the-shelf GPS as secondary information sources. Localization is performed by a particle filter approach, coupled with an Kalman filter for robustness and sensor fusion. This leads to a lateral accuracy below 20 cm in various urban test areas. The system has been included in our autonomous test vehicle and successfully demonstrated the full loop from mapping to autonomous driving.

Reliable autonomous navigation is still a challenging problem for robots with simple and inexpensive hardware. A key difficulty is the need to maintain an internal map of the environment and an accurate estimate of the robot's position in this map. Recently, a stigmergic approach has been proposed in which a navigation map is stored into the environment, on a grid of RFID tags, and robots use it to optimally reach predefined goal points without the need for internal maps. While effective, this approach is limited to a predefined set of goal points. In this paper, we extend this approach to enable robots to travel to any point on the RFID floor, even if it was not previously identified as a goal location, as well as to keep a safe distance from any given critical location. Our approach produces safe, repeatable and quasi-optimal trajectories without the use of internal maps, self localization, or path planning. We report experiments run in a real apartment equipped with an RFID floor, in which a service robot either reaches or avoids a user who wears slippers equipped with an RFID tag reader.

The objective of this paper is to investigate the problem of how to best combine and fuse color and depth measurements for incremental pose estimation or 3D tracking. Subsequently a framework will be proposed that allows to formulate the problem with a unique measurement vector and not to combine them in an ad-hoc manner. In particular, the full color and depth measurement will be defined as a 4-vector (by combining 3D Euclidean points + image intensities) and an optimal error for pose estimation will be derived from this. As will be shown, this will lead to designing an iterative closest point approach in 4 dimensional space. A kd-tree is used to find the closest point in 4D-space, therefore simultaneously accounting for color and depth. Based on this unified framework a novel Point-to-hyperplane approach will be introduced which has the advantages of classic Point-to-plane ICP but in 4D-space. By doing this it will be shown that there is no longer any need to provide or estimate a scale factor between different measurement types. Consequently, this allows to increase the convergence domain and speed up the alignment, whilst maintaining the robust and accurate properties. Results on both simulated and real environments will be provided along with benchmark comparisons.

Although Point Clouds Registration is a very well studied problem, with many different solutions, most of the approaches in the literature aims at aligning two dense point clouds. Instead, we tackle the problem of aligning a dense point cloud with a sparse one: a problem that has to be solved, for example, to merge maps produced by different sensors, such as a vision-based sensor and laser scanner or two different laser-based sensors. The most used approach to point clouds registration, Iterative Closest Point (ICP), is also applicable to this sub-problem. We propose an improvement over the standard ICP data association policy and we called it Probabilistic Data Association. It was derived applying statistical inference techniques on a fully probabilistic model. In our proposal, each point in the source point cloud is associated with a set of points in the target point cloud; each association is then weighted so that the weights form a probability distribution. The result is an algorithm similar to ICP but more robust w.r.t. noise and outliers. While we designed our approach to deal with the problem of dense-sparse registration, it can be successfully applied also to standard point clouds registration.

Our work addresses the problem of planning an optimal pitching trajectory for compliant serial manipulators. We propose a generic framework for finding the longest possible pitch given angle, velocity, and torque constraints. Within this framework we can decompose an original infinite-dimensional task to finite number of similar steps. As a result, the implemented procedure gives an optimal release configuration, temporal control torque profiles satisfying dynamic constraints, and optimal stiffness coefficients for compliant joints. Computational complexity of the nonlinear optimization task was reduced applying a novel motion parametrization method. The proposed method was illustrated for a particular example and experimentally verified with the KUKA LWR4+ arm.

We present a framework for planning collision-free and safe paths online for autonomous underwater vehicles (AUVs) in unknown environments. We build up on our previous work and propose an improved approach. While preserving its main modules (mapping, planning and mission handler), the framework now considers motion constraints to plan feasible paths, i.e., those that meet vehicle's motion capabilities. The new framework also incorporates a risk function to avoid navigating close to nearby obstacles, and reuses the last best known solution to eliminate time-consuming pruning routines. To evaluate this approach, we use the Sparus II AUV, a torpedo-shaped vehicle performing autonomous missions in a 2-dimensional workspace. We validate the framework's new features by solving tasks in both simulation and real-world in-water trials and comparing results with our previous approach.

the principle and control method of torso orientation in dynamic biped walking is addressed. First, the role of torso motion in walking performance, especially at the landing moment, is investigated by using a simplified model. Second, a control strategy of torso orientation is proposed, including the algorithms in the push-off and the single support phases (SSP). Furthermore, the module of torso control is integrated into the control method of dynamic biped walking, which was proposed in our previous research. The paper ends with the simulation of a humanoid robot with 21 internal degrees of freedom (DOF), and shows the feasibility of the proposed method.

The ability to perform an accurate robotic fold is essential to obtain the properly folded garment. Available solutions rely on a rough folding surface or on a comprehensive simulation, both preventing the garment from slipping on the table during folding. This paper proposes a new algorithm for a folding path design respecting the garment material properties and preventing the garment slipping. The folding path is derived based on the equilibrium of forces under the simplifying assumptions of a rectangular and homogeneous garment. This approach allows folding the rectangular garment on a low friction table surface as we demonstrated in the experiments performed by a dual-arm robotic testbed.

We present a damage-aware planning approach which determines the best sequence to manipulate a number of objects in a scene. This works on task-planning level, abstracts from motion planning and anticipates the dynamics of the scene using a physics simulation. Instead of avoiding interaction with the environment, we take unintended motion of other objects into account and plan manipulation sequences which minimize the potential damage. Our method can also be used as a validation measure to judge planned motions for their feasibility in terms of damage avoidance. We evaluate our approach on one industrial scenario (autonomous container unloading) and one retail scenario (shelf replenishment).

This paper describes a person identification method using a smartphone and laser range finders (LRFs) for a mobile service robot. The robot is equipped with LRFs and the target person holds a smartphone. The method first detects the foot strike timings of the target person using the smartphone and those of all people by using the LRFs. By finding the person whose foot strike timings captured by the LRFs are similar to those obtained by the smartphone, the robot can identify the target person. Person identification experiments and person following experiments are conducted in order to validate the method. Since the method only requires a person to simply hold a smartphone, it can be easily applied to daily situations.

This paper presents an online optimization-based approach to compute trajectories to enable substitution of robots in formation-based deployments with durations that exceed the energy capacity of individual systems. The proposed algorithm computes trajectories in a multi-robot context to ensure a collision-free exchange, even where congestion is a concern. The quality of the resulting trajectories is determined by the amount of time spent deviating from the original plan while maintaining collision-free, speed-limited polynomial splines. The algorithm is shown through simulation and experiments to be viable with average deviation time gaps of less than 16 seconds and average computation times of under 3 minutes for the presented scenarios with varying numbers of robots and deployment specifications.

We present a lifelong mapping and localisation system for long-term autonomous operation of mobile robots in changing environments. The core of the system is a spatio-temporal occupancy grid that explicitly represents the persistence and periodicity of the individual cells and can predict the probability of their occupancy in the future. During navigation, our robot builds temporally local maps and integrates then into the global spatio-temporal grid. Through re-observation of the same locations, the spatio-temporal grid learns the long-term environment dynamics and gains the ability to predict the future environment states. This predictive ability allows to generate time-specific 2d maps used by the robot's localisation and planning modules. By analysing data from a long-term deployment of the robot in a human-populated environment, we show that the proposed representation improves localisation accuracy and the efficiency of path planning. We also show how to integrate the method into the ROS navigation stack for use by other roboticists.