ICRA 2020 论文列表

Visual place recognition algorithms trade off three key characteristics: their storage footprint, their computational requirements, and their resultant performance, often expressed in terms of recall rate. Significant prior work has investigated highly compact place representations, sub-linear computational scaling and sub-linear storage scaling techniques, but have always involved a significant compromise in one or more of these regards, and have only been demonstrated on relatively small datasets. In this paper we present a novel place recognition system which enables for the first time the combination of ultra-compact place representations, near sub-linear storage scaling and extremely lightweight compute requirements. Our approach exploits the inherently sequential nature of much spatial data in the robotics domain and inverts the typical target criteria, through intentionally coarse scalar quantization-based hashing that leads to more collisions but is resolved by sequence-based matching. For the first time, we show how effective place recognition rates can be achieved on a new very large 10 million place dataset, requiring only 8 bytes of storage per place and 37K unitary operations to achieve over 50% recall for matching a sequence of 100 frames, where a conventional stateof-the-art approach both consumes 1300 times more compute and fails catastrophically. We present analysis investigating the effectiveness of our hashing overload approach under varying sizes of quantized vector length, comparison of near miss matches with the actual match selections and characterise the effect of variance re-scaling of data on quantization. Resource link: https://github.com/oravus/CoarseHash.

This paper presents a task and motion planning (TAMP) framework for a robotic manipulator in order to retrieve a target object from clutter. We consider a configuration of objects in a confined space with a high density so no collision-free path to the target exists. The robot must relocate some objects to retrieve the target without collisions. For fast completion of object rearrangement, the robot aims to optimize the number of pick-and-place actions which often determines the efficiency of a TAMP framework.We propose a task planner incorporating motion planning to generate executable plans which aims to minimize the number of pick-and-place actions. In addition to fully known and static environments, our method can deal with uncertain and dynamic situations incurred by occluded views. Our method is shown to reduce the number of pick-and-place actions compared to baseline methods (e.g., at least 28.0% of reduction in a known static environment with 20 objects).

Surgery may involve precise instrument tip positioning in a minimally invasive way. During these operations, the instrument is inserted in the body through an orifice. The movements of the instrument are constrained by interaction forces arising at the orifice level. The physical constraints may drastically vary depending on the patient’s anatomy. This introduces uncertainties that challenge the positioning task for a robot. Indeed, it raises an antagonism: On one side, the required precision appeals for a rigid behavior. On the other side, forces applied at the entry point should be limited, which requires softness. In this paper we choose to minimize forces at the orifice by using a passive ball joint wrist to manipulate the instrument. From a control perspective, this leads to consider the task as a 3 DOF wrist center positioning problem, whose softness can be achieved through conventional low impedance control. However, positioning the wrist center, even with a high static precision, does not allow to achieve a high precision of the instrument tip positioning when the orifice behavior is not known. To cope with this problem, we implement a controller that servos the tip position by commanding the wrist position. In order to deal with uncertainties, we exploit an adaptive control scheme that identifies in real-time the unknown mapping between the wrist velocity and the tip velocity. Both simulations and in vitro experimental results show the efficiency of the control law.

This paper considers the problem of fast and safe autonomous navigation in partially known environments. Our main contribution is a control policy design based on ellipsoidal trajectory bounds obtained from a quadratic state-dependent distance metric. The ellipsoidal bounds are used to embed directional preference in the control design, leading to system behavior that is adapted to local environment geometry, carefully considering medial obstacles while paying less attention to lateral ones. We use a virtual reference governor system to adaptively follow a desired navigation path, slowing down when system safety may be violated and speeding up otherwise. The resulting controller is able to navigate complex environments faster than common Euclidean-norm and Lyapunov-function-based designs, while retaining stability and collision avoidance guarantees.

We introduce a new, analytical method for generating whole-body motions for humanoid robots, which approximate the desired Composite Rigid Body (CRB) inertia. Our approach uses a reduced five mass model, where four of the masses are attributed to the limbs and one is used for the trunk. This compact formulation allows for finding an analytical solution that combines the kinematics with mass distribution and inertial properties of a humanoid robot. The positioning of the masses in Cartesian space is then directly used to obtain joint angles with relations based on simple geometry. Motions are achieved through the time evolution of poses generated through the desired foot positioning and CRB inertia properties. As a result, we achieve short computation times in the order of tens of microseconds. This makes the method suited for applications with limited computation resources, or leaving them to be spent on higher-layer tasks such as model predictive control. The approach is evaluated by performing a dynamic kicking motion with an igus® Humanoid Open Platform robot.

In this paper, we present a fast, on-line mapping and planning solution for operation in unknown, off-road, environments. We combine obstacle detection along with a terrain gradient map to make simple and adaptable cost map. This map can be created and updated at 10 Hz. An A* planner finds optimal paths over the map. Finally, we take multiple samples over the control input space and do a kinematic forward simulation to generated feasible trajectories. Then the most optimal trajectory, as determined by the cost map and proximity to A* path, is chosen and sent to the controller. Our method allows real time operation at rates of 30 Hz. We demonstrate the efficiency of our method in various off-road terrain at high speed.

Exploration and collision-free navigation through an unknown environment is a fundamental task for autonomous robots. In this paper, a novel exploration strategy for Micro Aerial Vehicles (MAVs) is presented. The goal of the exploration strategy is the reduction of map entropy regarding occupancy probabilities, which is reflected in a utility function to be maximised. We achieve fast and efficient exploration performance with tight integration between our octree-based occupancy mapping approach, frontier extraction, and motion planning-as a hybrid between frontier-based and sampling-based exploration methods. The computationally expensive frontier clustering employed in classic frontier-based exploration is avoided by exploiting the implicit grouping of frontier voxels in the underlying octree map representation. Candidate next-views are sampled from the map frontiers and are evaluated using a utility function combining map entropy and travel time, where the former is computed efficiently using sparse raycasting. These optimisations along with the targeted exploration of frontier-based methods result in a fast and computationally efficient exploration planner. The proposed method is evaluated using both simulated and real-world experiments, demonstrating clear advantages over state-of-the-art approaches.

Deep reinforcement learning (RL) is being actively studied for robot navigation due to its promise of superior performance and robustness. However, most existing deep RL navigation agents are trained using fixed parameters, such as maximum velocities and weightings of reward components. Since the optimal choice of parameters depends on the use-case, it can be difficult to deploy such existing methods in a variety of real-world service scenarios. In this paper, we propose a novel deep RL navigation method that can adapt its policy to a wide range of parameters and reward functions without expensive retraining. Additionally, we explore a Bayesian deep learning method to optimize these parameters that requires only a small amount of preference data. We empirically show that our method can learn diverse navigation skills and quickly adapt its policy to a given performance metric or to human preference. We also demonstrate our method in real-world scenarios.

Real-time semantic segmentation is desirable in many robotic applications with limited computation resources. One challenge of semantic segmentation is to deal with the object scale variations and leverage the context. How to perform multi-scale context aggregation within limited computation budget is important. In this paper, firstly, we introduce a novel and efficient module called Cascaded Factorized Atrous Spatial Pyramid Pooling (CF-ASPP). It is a lightweight cas-caded structure for Convolutional Neural Networks (CNNs) to efficiently leverage context information. On the other hand, for runtime efficiency, state-of-the-art methods will quickly decrease the spatial size of the inputs or feature maps in the early network stages. The final high-resolution result is usually obtained by non-parametric up-sampling operation (e.g. bilinear interpolation). Differently, we rethink this pipeline and treat it as a super-resolution process. We use optimized super-resolution operation in the up-sampling step and improve the accuracy, especially in sub-sampled input image scenario for real-time applications. By fusing the above two improvements, our methods provide better latency-accuracy trade-off than the other state-of-the-art methods. In particular, we achieve 68.4% mIoU at 84 fps on the Cityscapes test set with a single Nivida Titan X (Maxwell) GPU card. The proposed module can be plugged into any feature extraction CNN and benefits from the CNN structure development.

Interactive behavior is important for autonomous driving vehicles, especially for scenarios like ramp merging which require significant social interaction between autonomous driving vehicles and human-driven cars. This paper enhances our previous Probabilistic Graphical Model (PGM) merging control model for the interactive behavior of autonomous driving vehicles. To better estimate the interactive behavior for autonomous driving cars, a Factor Graph (FG) is used to describe the dependency among observations and estimate other cars’ intentions. Real trajectories are used to approximate the model instead of human-designed models or cost functions. Forgetting factors and a Gaussian Mixture Model (GMM) are also applied in the intention estimation process for stabilization, interpolation and smoothness. The advantage of the factor graph is that the relationship between its nodes can be described by self-defined functions, instead of probabilistic relationships as in PGM, giving more flexibility. Continuity of GMM also provides higher accuracy than the previous discrete speed transition model. The proposed method enhances the overall performance of intention estimation, in terms of collision rate and average distance between cars after merging, which means it is safer and more efficient.

Accurate force sensing is important for endoluminal intervention in terms of both safety and lesion targeting. This paper develops an FBG-based force sensor for robotic bronchoscopy by configuring three FBG sensors at the lateral side of a conical substrate. It allows a large and eccentric inner lumen for the interventional instrument, enabling a flexible imaging probe inside to perform optical biopsy. The force sensor is embodied with a laser-profiled continuum robot and thermo drift is fully compensated by three temperature sensors integrated on the circumference surface of the sensor substrate. Different decoupling approaches are investigated, and nonlinear decoupling is adopted based on the cross-validation SVM and a Gaussian kernel function, achieving an accuracy of 10.58 mN, 14.57 mN and 26.32 mN along X, Y and Z axis, respectively. The tissue test is also investigated to further demonstrate the feasibility of the developed triaxial force sensor.

Deep neural networks (DNNs) have achieved great success in the area of computer vision. The disparity estimation problem tends to be addressed by DNNs which achieve much better prediction accuracy in stereo matching than traditional hand-crafted feature based methods. On one hand, however, the designed DNNs require significant memory and computation resources to accurately predict the disparity, especially for those 3D convolution based networks, which makes it difficult for deployment in real-time applications. On the other hand, existing computation-efficient networks lack expression capability in large-scale datasets so that they cannot make an accurate prediction in many scenarios. To this end, we propose an efficient and accurate deep network for disparity estimation named FADNet with three main features: 1) It exploits efficient 2D based correlation layers with stacked blocks to preserve fast computation; 2) It combines the residual structures to make the deeper model easier to learn; 3) It contains multi-scale predictions so as to exploit a multi-scale weight scheduling training technique to improve the accuracy. We conduct experiments to demonstrate the effectiveness of FADNet on two popular datasets, Scene Flow and KITTI 2015. Experimental results show that FADNet achieves state-of-the-art prediction accuracy, and runs at a significant order of magnitude faster speed than existing 3D models. The codes of FADNet are available at https://github.com/HKBU-HPML/FADNet.

Magnetic helical microswimmers can be propelled by rotating magnetic field and are adept at passing through narrow space. To date, various magnetic actuation systems and control methods have been developed to drive these microswimmers. However, steering their spacial movement in a large workspace is still challenging, which could be significant for potential medical applications. In this regard, this paper designs an eye-in-hand stereo-vision module and corresponding refraction-rectified location algorithm. Combined with the motor module and the coil module, the mobile-coil system is capable of generating dynamic magnetic fields in a large 3D workspace. Based on the system, a robust triple-loop stereo visual servoing strategy is proposed that operates simultaneous tracking, locating, and steering, through which the helical swimmer is able to follow a long-distance 3D path. A scaled-up magnetic helical swimmer is employed in the path following experiment. Our prototype system reaches a cylindrical workspace with a diameter more than 200 mm, and the mean error of path tracking is less than 2 mm.

Optimization based predictive control is a powerful tool that has improved the ability of legged robots to execute dynamic maneuvers and traverse increasingly difficult terrains. However, it is often challenging and unintuitive to design meaningful cost functions and build high-fidelity models while adhering to timing restrictions. A novel framework to extract and design principled regularization heuristics for legged locomotion optimization control is presented. By allowing a simulation to fully explore the cost space offline, certain states and actions can be constrained or isolated. Data is fit with simple models relating the desired commands, optimal control actions, and robot states to identify new heuristic candidates. Basic parameter learning and adaptation laws are then applied to the models online. This method extracts simple, but powerful heuristics that can approximate complex dynamics and account for errors stemming from model simplifications and parameter uncertainty without the loss of physical intuition while generalizing the parameter tuning process. Results on the Mini Cheetah robot verify the increased capabilities due to the newly extracted heuristics without any modification to the controller structure or gains.

Riemannian Motion Policies (RMPs) have recently been introduced as a way to specify second-order motion policies defined on robot task spaces. RMP-based approaches have the advantage of being more general than traditional approaches based on operational space control; for example, the generalized task inertia in an RMP can be fully state-dependent, which is particularly effective in designing collision avoidance bahaviors. But until now RMPs have been applied only to fully actuated systems, i.e. systems for which each degree of freedom (DoF) can be directly actuated by a control input. In this paper, we present a method that extends the RMP formalism to a class of underacutated systems whose dynamics are amenable to a decomposition into a fully-actuated subsystem and a residual dynamics. We show the efficacy of the approach by constructing a suitable decomposition for a Wheeled-Inverted-Pendulum (WIP) humanoid robot and applying our method to derive motion policies for combined locomotion and manipulation tasks. Simulation results are presented for a 7-DoF system with one degree of underactuation.

Mobile robots can be used in numerous outdoor tasks such as patrolling, delivery and military applications. In order to deploy mobile robots in this kind of environment, where there are different challenges like slopes, elevations, or even holes, they should be able to detect such challenges and determine the best path to accomplish their tasks. In this paper, we are proposing an exploration approach based on potential fields with local distortions, in which we define preferences in uneven terrains to avoid high declivity regions without compromising the best path. The approach was implemented and tested in simulated environments, considering a ground robot embedded with two 2D LIDAR sensors, and the experiments demonstrated the efficiency of our method.

This paper presents a robot trajectory optimization formulation that builds upon numerical optimal control and Lie group methods. In particular, the inherent sparsity of direct collocation is carefully analyzed to dramatically reduce the number of floating-point operations to get first-order information of the problem. We describe how sparsity exploitation is employed with both numerical and analytical differentiation. Furthermore, the use of geometric algorithms based on Lie groups and their associated Lie algebras allow to analytically evaluate the state equations and their derivatives with efficient recursive algorithms. We demonstrate the scalability of the proposed formulation with three different articulated robots, such as a finger, a mobile manipulator and a humanoid composed of five, eight and more than twenty degrees of freedom, respectively. The performance of our implementation in C++ is also validated and compared against a state-of-the-art general purpose numerical optimal control solver.

Over the past decade we have been able to vastly improve the control algorithms of our biped walking robot LOLA. Further enhancements, however, are limited by vibration problems caused by the dynamics of LOLA's mechanical structure. In this work, we present small examples how structural dynamics limit our control design for walking control as well as low level position control of the joints. We also provide a procedure to identify weaknesses in the structural design of our biped using Experimental Modal Analysis. Using this method, we could successfully identify the structural modes of the system. Furthermore, we were able to use a closed-loop identification method to show a connection between the control loop resonances and the structural resonances of our robot.

In the robotics literature, different knowledge transfer approaches have been proposed to leverage the experience from a source task or robot-real or virtual-to accelerate the learning process on a new task or robot. A commonly made but infrequently examined assumption is that incorporating experience from a source task or robot will be beneficial. In practice, inappropriate knowledge transfer can result in negative transfer or unsafe behaviour. In this work, inspired by a system gap metric from robust control theory, the ν-gap, we present a data-efficient algorithm for estimating the similarity between pairs of robot systems. In a multi-source inter-robot transfer learning setup, we show that this similarity metric allows us to predict relative transfer performance and thus informatively select experiences from a source robot before knowledge transfer. We demonstrate our approach with quadrotor experiments, where we transfer an inverse dynamics model from a real or virtual source quadrotor to enhance the tracking performance of a target quadrotor on arbitrary hand-drawn trajectories. We show that selecting experiences based on the proposed similarity metric effectively facilitates the learning of the target quadrotor, improving performance by 62% compared to a poorly selected experience.