IROS 2017 Papers

Autonomous navigation, which consists of a systematic integration of localization, mapping, motion planning and control, is the core capability of mobile robotic systems. However, most research considers only isolated technical modules. There exist significant gaps between maps generated by SLAM algorithms and maps required for motion planning. This paper presents a complete online system that consists in three modules: incremental SLAM, real-time dense mapping, and free space extraction. The obtained free-space volume (i.e. a tessellation of tetrahedra) can be served as regular geometric constraints for motion planning. Our system runs in real-time thanks to the engineering decisions proposed to increase the system efficiency. We conduct extensive experiments on the KITTI dataset to demonstrate the run-time performance. Qualitative and quantitative results on mapping accuracy are also shown. For the benefit of the community, we make the source code public.

BRIDGE bot is a 158 g, 10.7 × 8.9 × 6.5 cm3, magnetic-wheeled robot designed to traverse and inspect steel bridges. Utilizing custom magnetic wheels, the robot is able to securely adhere to the bridge in any orientation. The body platform features flexible, multi-material legs that enable a variety of plane transitions as well as robot shape manipulation. The robot is equipped with a Cortex-M0 processor, inertial sensors, and a modular wireless radio. A camera is included to provide images for detection and evaluation of identified problems. The robot has been demonstrated moving through plane transitions from 45° to 340° as well as over obstacles up to 9.5 mm in height. Preliminary use of sensor feedback to improve plane transitions has also been demonstrated.

Surgical interventions are increasingly executed minimal invasively. Surgeons insert instruments through tiny incisions in the body and pivot slender instruments to treat organs or tissue below the surface. While a blessing for patients, surgeons need to pay extra attention to overcome the fulcrum effect, reduced haptic feedback and deal with lost hand-eye coordination. The mental load makes it difficult to pay sufficient attention to the forces that are exerted on the body wall. In delicate procedures such as fetal surgery, this might be problematic as irreparable damage could cause premature delivery. As a first attempt to quantify the interaction forces applied on the patient's body wall, a novel 6 degrees of freedom force sensor was developed for an ex-vivo set up. The performance of the sensor was characterised. User experiments were conducted by 3 clinicians on a set up simulating a fetal surgical intervention. During these simulated interventions, the interaction forces were recorded and analysed when a normal instrument was employed. These results were compared with a session where a flexible instrument under haptic guidance was used. The conducted experiments resulted in interesting insights in the interaction forces and stresses that develop during such difficult surgical intervention. The results also implicated that haptic guidance schemes and the use of flexible instruments rather than rigid ones could have a significant impact on the stresses that occur at the body wall.

The most data-efficient algorithms for reinforcement learning (RL) in robotics are based on uncertain dynamical models: after each episode, they first learn a dynamical model of the robot, then they use an optimization algorithm to find a policy that maximizes the expected return given the model and its uncertainties. It is often believed that this optimization can be tractable only if analytical, gradient-based algorithms are used; however, these algorithms require using specific families of reward functions and policies, which greatly limits the flexibility of the overall approach. In this paper, we introduce a novel model-based RL algorithm, called Black-DROPS (Black-box Data-efficient RObot Policy Search) that: (1) does not impose any constraint on the reward function or the policy (they are treated as black-boxes), (2) is as data-efficient as the state-of-the-art algorithm for data-efficient RL in robotics, and (3) is as fast (or faster) than analytical approaches when several cores are available. The key idea is to replace the gradient-based optimization algorithm with a parallel, black-box algorithm that takes into account the model uncertainties. We demonstrate the performance of our new algorithm on two standard control benchmark problems (in simulation) and a low-cost robotic manipulator (with a real robot).

For humanoid robots, unexpected collision can cause instability of robot balancing and damage to both robots and environment. This paper presents a reactive bipedal walking controller against swing foot collision for humanoid robots. This controller is composed of following three components: 1) Swing Foot Trajectory Regenerator, 2) Swing Foot Collision Detector, and 3) Swing Foot Impact Mitigation Controller. By regenerating swing foot trajectory depending on situations, humanoid robots can avoid falling down. However, although humanoid robots detect collision and regenerate a swing foot, collision impact can cause bad effects such as damage and posture rotation. Therefore, to mitigate strong impact, we propose Swing Foot Impact Mitigation Controller, which is composed of two controllers. The proposed method is validated through the experiments by actual humanoid robot CHIDORI. We confirm that CHIDORI can avoid falling down against collision in two situations: walking on the flat ground, and stepping up a stair.

In this study, we propose a novel method to operate whole body of a humanoid robot, which also includes both feet, dynamically and safely with the master-slave approach. The conventional whole body master-slave approaches need static balancing assumption or a certain time length of planning after operator's input. Then, we introduce a set of limitations that allows the robot to execute human's daily dynamic bipedal locomotion, but forbid dangerous motions like the COM will be gone outside of the support region. In the limitations, we regulate COM velocity based on a positional relation of the Divergent Component of Motion (Capture Point) and the both feet, and automatically modify the swing foot contact timing with judging the ZMP is inside or outside of the single foot support region. At last, we conducted some experiments of the real time master-slave locomotion with using two life-sized humanoid robots and confirmed the effectiveness of our novel limitation methods.

In this paper we introduce a biped foot placement controller based on capture point (CP) feedback control. This capture point feedback controller generates desired Zero Moment Point (ZMP) according to current capture point error. We call this desired ZMP as Control-ZMP (cZMP). Using this cZMP, we constructed (i) a disturbance adapting walking pattern generator, (ii) an ankle torque reference generator, (iii) a landing foot position adjustor and, (iv) a step time adjustor. By applying these controllers, the biped system becomes robust over uneven terrain and external pushing disturbances. We considered cZMP as a key indicator of bipedal walking stability. If the cZMP is within the support polygon, the robot will track the cZMP by using (i) and (ii). If the cZMP is outside of the support polygon, the robot will change footstep position and time by using (iii) and (iv) (COM pattern will be changed by (i) too). Capture point dynamics is described based on a linear inverted pendulum (LIP) model. The performance of the suggested control algorithm is validated in the Choreonoid simulator with a model of DRC-HUBO+ [11], [12].

This paper describes a method of automatic sensor bias calculation for angle only sensor models in a target tracking scenario. It is assumed that separate Kalman filters are employed by each sensor and no measurements of known landmarks are available. Accurate bias estimation is achieved through the use of pseudo measurements of slant range from each sensor to the target and pseudo measurements of each sensor's bias. The covariance intersection (CI) algorithm is used to produce a pseudo measurement of slant range. This pseudo measurement of range allows pseudo measurements of sensor bias to be calculated based on each sensor's residuals and Kalman gains. Substantially improved tracking performance is demonstrated when estimating and accounting for constant biases on each sensor.

In this paper a binary feature based Loop Closure Detection (LCD) method is proposed, which for the first time achieves higher precision-recall (PR) performance compared with state-of-the-art SIFT feature based approaches. The proposed system originates from our previous work Multi-Index hashing for Loop closure Detection (MILD), which employs Multi-Index Hashing (MIH) [1] for Approximate Nearest Neighbor (ANN) search of binary features. As the accuracy of MILD is limited by repeating textures and inaccurate image similarity measurement, burstiness handling is introduced to solve this problem and achieves considerable accuracy improvement. Additionally, a comprehensive theoretical analysis on MIH used in MILD is conducted to further explore the potentials of hashing methods for ANN search of binary features from probabilistic perspective. This analysis provides more freedom on best parameter choosing in MIH for different application scenarios. Experiments on popular public datasets show that the proposed approach achieved the highest accuracy compared with state-of-the-art while running at 30Hz for databases containing thousands of images.

Active Object Recognition (AOR) has been approached as an unsupervised learning problem, in which optimal trajectories for object inspection are not known and to be discovered by reducing label uncertainty or training with reinforcement learning. Such approaches suffer from local optima and have no guarantees of the quality of their solution. In this paper, we treat AOR as a Partially Observable Markov Decision Process (POMDP) and find near-optimal values and corresponding action-values of training data using Belief Tree Search (BTS) on the AOR belief Markov Decision Process (MDP). AOR then reduces to the problem of knowledge transfer from these action-values to the test set. We train a Long Short Term Memory (LSTM) network on these values to predict the best next action on the training set rollouts and experimentally show that our method generalizes well to explore novel objects and novel views of familiar objects with high accuracy. We compare this supervised scheme against guided policy search, and show that the LSTM network reaches higher recognition accuracy compared to the guided policy search and guided Neurally Fitted Q-iteration. We further look into optimizing the observation function to increase the total collected reward during active recognition. In AOR, the observation function is known only approximately. We derive a gradient-based update for the observation function to increase the total expected reward. We show that by optimizing the observation function and retraining the supervised LSTM network, the AOR performance on the test set improves significantly.

We characterize and propose advances in the technique of Belief Roadmap Search (BRMS), the process of searching a roadmap in belief space for robot motion planning under localization uncertainty. We discuss the conditions required for optimal substructure in the single-source search of a roadmap in belief space, demonstrating that there are several desirable cost functions for which this property cannot be achieved. Practical performance issues of BRMS are discussed, including the implications of a commonly-used anti-cycling rule, and the computational complexity realized in practical applications of the technique. We propose a best-first implementation of BRMS, in contrast to the standard breadth-first implementation, which we show to improve the computational cost of search by up to 49% by eliminating unnecessary node expansions - the mechanics of both approaches are compared in detail. A variety of motion planning examples are explored.

Designing and representing control algorithms is challenging in swarm robotics, where the collective swarm performance depends on interactions between robots and with their environment. The currently available modeling languages, such as UML, cannot fully express these interactions. We therefore propose a new, Behaviour-Data Relations Modeling Language (BDRML), where robot behaviours and data that robots utilise, as well as relationships between them, are explicitly represented. This allows BDRML to express control algorithms where robots cooperate and share information with each other while interacting with the environment.

For fast-moving legged robots, the ability to stop rapidly is essential if they are to move safely through unpredictable environments. There is a trade-off between stability and rapidity however, as the forces that slow the robot down also tend to cause a toppling moment. This paper examines how rapid deceleration motions performed on a bipedal robot are affected if it is forced to maintain zero rate of change in angular momentum (ZRAM) throughout the maneuver. By using trajectory optimization to generate over 4000 gait termination motions from speeds comparable to a sprinting human, it was found that the stopping distance could be reduced by 12% and time by 25% by relaxing the ZRAM condition, with greater reductions possible through the addition of a free stabilizing limb. The ability to moderate the pitch of the robot while decelerating was found to be paramount to achieving a rapid stop and thus, a template incorporating an inertial body and stabilizing appendages is proposed to represent these maneuvers.

Camera relocalization plays a vital role in many robotics and computer vision tasks, such as global localization, recovery from tracking failure, and loop closure detection. Recent random forests based methods directly predict 3D world locations for 2D image locations to guide the camera pose optimization. During training, each tree greedily splits the samples to minimize the spatial variance. However, these greedy splits often produce uneven sub-trees in training or incorrect 2D-3D correspondences in testing. To address these problems, we propose a sample-balanced objective to encourage equal numbers of samples in the left and right sub-trees, and a novel backtracking scheme to remedy the incorrect 2D-3D correspondence predictions. Furthermore, we extend the regression forests based methods to use local features in both training and testing stages for outdoor RGB-only applications. Experimental results on publicly available indoor and outdoor datasets demonstrate the efficacy of our approach, which shows superior or on-par accuracy with several state-of-the-art methods.

In this paper, a current control scheme for the brushless DC (BLDC) motor is proposed. In order to estimate the phase current, it employs the adaptive extended Kalman filter (EKF) and measurement of a single DC link current. When estimating the motor phase current, measurement of a single DC link current is scaled by PWM (Pulse Width Modulation) duty. If the PWM duty is very small, the sensitivity of the measurement noise is dramatically increased. In order to solve this problem adaptive EKF is applied. Therefore, estimated phase current using adaptive EKF prevents the possible occurrence of an unstable situation due to the sensitivity. The accuracy of the proposed current control scheme was confirmed by comparing the estimated current by adaptive EKF and the measured actual phase current. In additions, the effectiveness was verified through the current-control time response of the when the PWM duty for the control was close to zero. By using the proposed current control scheme, robust current control of the BLDC motor was possible with a single DC link current.

To increase the autonomy of humanoid robots, the visual perception must support the efficient collection and interpretation of visual scene cues by providing task-dependent information. Active vision systems allow to extend the observable workspace by employing active gaze control, i.e. by shifting the gaze to relevant areas in the scene. When moving the eyes, stabilization of the camera images is crucial for successful task execution. In this paper, we present an active vision system for task-oriented selection of view directions and gaze stabilization to enable a humanoid robot to robustly perform vision-based tasks. We investigate the interaction between a gaze stabilization controller and view planning to select the next best view direction based on saliency maps which encode task-relevant information. We demonstrate the performance of the systems in a real world scenario, in which a humanoid robot is performing vision-based grasping while moving, a task that would not be possible without the combination of view selection and gaze stabilization.

Software testing is an important tool to ensure software quality. This is a hard task in robotics due to dynamic environments and the expensive development and time-consuming execution of test cases. Most testing approaches use model-based and/or simulation-based testing to overcome these problems. We propose model-free skill-centric testing in which a robot autonomously executes skills in the real world and compares it to previous experiences. The skills are selected by maximising the expected information gain on the distribution of erroneous software functions. We use deep learning to model the sensor data observed during previous successful skill executions and to detect irregularities. Sensor data is connected to function call profiles such that certain misbehaviour can be related to specific functions. We evaluate our approach in simulation and in experiments with a KUKA LWR 4+ robot by purposefully introducing bugs to the software. We demonstrate that these bugs can be detected with high accuracy and without the need for the implementation of specific tests or task-specific models.

Efficient robotic navigation requires a predefined map. Various autonomous exploration strategies exist, which direct robots to unexplored space by detecting frontiers. Frontiers are boundaries separating known space form unknown space. Usually frontier detection utilizes image processing tools like edge detection, thus limiting it to two dimensional (2D) exploration. This paper presents a new exploration strategy based on the use of multiple Rapidly-exploring Random Trees (RRTs). The RRT algorithm is chosen because, it is biased towards unexplored regions. Also, using RRT provides a general approach which can be extended to higher dimensional spaces. The proposed strategy is implemented and tested using the Robot Operating System (ROS) framework. Additionally this work uses local and global trees for detecting frontier points, which enables efficient robotic exploration. Current efforts are limited to the single robot case. Extension to multi-agent systems and three-dimensional (3D) space is left for future effort.

This paper presents a model for safe driving at blind intersections and its integration to a local planner based on a Frenet frame. The model predicts potential moving obstacles from blind intersections to proactively slow down to avoid potential collisions. The derivation of the model is described and its parameters are detailed. The local planner computes smooth trajectories with smooth velocity profiles so that the vehicle can follow the paths without jerk and sudden accelerations resulting in safe and comfortable navigation. Experimental results in simulation and in the real field with an autonomous car, show that the proposed predictive driving framework can reproduce human expert driver's trajectories and velocities when facing blind intersections.