Spring-based actuators in legged locomotion provide energy-efficiency and improved performance, but increase the difficulty of controller design. Whereas previous works have focused on extensive modeling and simulation to find optimal controllers for such systems, we propose to learn model-free controllers directly on the real robot. In our approach, gaits are first synthesized by central pattern generators (CPGs), whose parameters are optimized to quickly obtain an open-loop controller that achieves efficient locomotion. Then, to make that controller more robust and further improve the performance, we use reinforcement learning to close the loop, to learn corrective actions on top of the CPGs. We evaluate the proposed approach in DLR's elastic quadruped bert. Our results in learning trotting and pronking gaits show that exploitation of the spring actuator dynamics emerges naturally from optimizing for dynamic motions, yielding high-performing locomotion despite being model-free. The whole process takes no more than 1.5 hours on the real robot and results in natural-looking gaits.
相關內容
In robotics motion is often described from an external perspective, i.e., we give information on the obstacle motion in a mathematical manner with respect to a specific (often inertial) reference frame. In the current work, we propose to describe the robotic motion with respect to the robot itself. Similar to how we give instructions to each other (go straight, and then after multiple meters move left, and then a sharp turn right.), we give the instructions to a robot as a relative rotation. We first introduce an obstacle avoidance framework that allows avoiding star-shaped obstacles while trying to stay close to an initial (linear or nonlinear) dynamical system. The framework of the local rotation is extended to motion learning. Automated clustering defines regions of local stability, for which the precise dynamics are individually learned. The framework has been applied to the LASA-handwriting dataset and shows promising results.
An intuitive control method for the flying trot, which combines offline trajectory planning with real-time balance control, is presented. The motion features of running animals in the vertical direction were analysed using the spring-load-inverted-pendulum (SLIP) model, and the foot trajectory of the robot was planned, so the robot could run similar to an animal capable of vertical flight, according to the given height and speed of the trunk. To improve the robustness of running, a posture control method based on a foot acceleration adjustment is proposed. A novel kinematic based CoM observation method and CoM regulation method is present to enhance the stability of locomotion. To reduce the impact force when the robot interacts with the environment, the virtual model control method is used in the control of the foot trajectory to achieve active compliance. By selecting the proper parameters for the virtual model, the oscillation motion of the virtual model and the planning motion of the support foot are synchronized to avoid the large disturbance caused by the oscillation motion of the virtual model in relation to the robot motion. The simulation and experiment using the quadruped robot Billy are reported. In the experiment, the maximum speed of the robot could reach 4.73 times the body length per second, which verified the feasibility of the control method.
Rapid development in deep learning model construction has prompted an increased need for appropriate training data. The popularity of large datasets - sometimes known as "big data" - has diverted attention from assessing their quality. Training on large datasets often requires excessive system resources and an infeasible amount of time. Furthermore, the supervised machine learning process has yet to be fully automated: for supervised learning, large datasets require more time for manually labeling samples. We propose a method of curating smaller datasets with comparable out-of-distribution model accuracy after an initial training session using an appropriate distribution of samples classified by how difficult it is for a model to learn from them.
Deep Reinforcement Learning (RL) is mainly studied in a setting where the training and the testing environments are similar. But in many practical applications, these environments may differ. For instance, in control systems, the robot(s) on which a policy is learned might differ from the robot(s) on which a policy will run. It can be caused by different internal factors (e.g., calibration issues, system attrition, defective modules) or also by external changes (e.g., weather conditions). There is a need to develop RL methods that generalize well to variations of the training conditions. In this article, we consider the simplest yet hard to tackle generalization setting where the test environment is unknown at train time, forcing the agent to adapt to the system's new dynamics. This online adaptation process can be computationally expensive (e.g., fine-tuning) and cannot rely on meta-RL techniques since there is just a single train environment. To do so, we propose an approach where we learn a subspace of policies within the parameter space. This subspace contains an infinite number of policies that are trained to solve the training environment while having different parameter values. As a consequence, two policies in that subspace process information differently and exhibit different behaviors when facing variations of the train environment. Our experiments carried out over a large variety of benchmarks compare our approach with baselines, including diversity-based methods. In comparison, our approach is simple to tune, does not need any extra component (e.g., discriminator) and learns policies able to gather a high reward on unseen environments.
Sim-to-real is a mainstream method to cope with the large number of trials needed by typical deep reinforcement learning. However, transferring a policy trained in simulation to actual hardware remains challenging due to the reality gap. In particular, the characteristics of actuators in legged robots have a considerable influence on sim-to-real transfer. High reduction ratio gears are widely used in actuators, and the reality gap issue becomes especially pronounced when even the utilization of backdrivability is considered to control joints compliantly. We propose a new simulation model of gears to address this gap. Additionally, the difficulty in achieving stable bipedal locomotion causes typical methods to fail to tune physical parameters in simulation with the behavior of transferred policy. Thus, we propose a method for system identification that can utilize failed attempts. The method's effectiveness is verified using a biped robot, the ROBOTIS-OP3, and the sim-to-real transferred policy can stabilize the robot under severe disturbances and walk on uneven surfaces without force and torque sensors.
Vision-based sensors have shown significant performance, accuracy, and efficiency gain in Simultaneous Localization and Mapping (SLAM) systems in recent years. In this regard, Visual Simultaneous Localization and Mapping (VSLAM) methods refer to the SLAM approaches that employ cameras for pose estimation and map generation. We can see many research works that demonstrated VSLAMs can outperform traditional methods, which rely only on a particular sensor, such as a Lidar, even with lower costs. VSLAM approaches utilize different camera types (e.g., monocular, stereo, and RGB-D), have been tested on various datasets (e.g., KITTI, TUM RGB-D, and EuRoC) and in dissimilar environments (e.g., indoors and outdoors), and employ multiple algorithms and methodologies to have a better understanding of the environment. The mentioned variations have made this topic popular for researchers and resulted in a wide range of VSLAMs methodologies. In this regard, the primary intent of this survey is to present the recent advances in VSLAM systems, along with discussing the existing challenges and trends. We have given an in-depth literature survey of forty-five impactful papers published in the domain of VSLAMs. We have classified these manuscripts by different characteristics, including the novelty domain, objectives, employed algorithms, and semantic level. We also discuss the current trends and future directions that may help researchers investigate them.
Collision-free motion generation in unknown environments is a core building block for robot manipulation. Generating such motions is challenging due to multiple objectives; not only should the solutions be optimal, the motion generator itself must be fast enough for real-time performance and reliable enough for practical deployment. A wide variety of methods have been proposed ranging from local controllers to global planners, often being combined to offset their shortcomings. We present an end-to-end neural model called Motion Policy Networks (M$\pi$Nets) to generate collision-free, smooth motion from just a single depth camera observation. M$\pi$Nets are trained on over 3 million motion planning problems in over 500,000 environments. Our experiments show that M$\pi$Nets are significantly faster than global planners while exhibiting the reactivity needed to deal with dynamic scenes. They are 46% better than prior neural planners and more robust than local control policies. Despite being only trained in simulation, M$\pi$Nets transfer well to the real robot with noisy partial point clouds. Code and data are publicly available at //mpinets.github.io.
Motion planning and control are crucial components of robotics applications. Here, spatio-temporal hard constraints like system dynamics and safety boundaries (e.g., obstacles in automated driving) restrict the robot's motions. Direct methods from optimal control solve a constrained optimization problem. However, in many applications finding a proper cost function is inherently difficult because of the weighting of partially conflicting objectives. On the other hand, Imitation Learning (IL) methods such as Behavior Cloning (BC) provide a intuitive framework for learning decision-making from offline demonstrations and constitute a promising avenue for planning and control in complex robot applications. Prior work primarily relied on soft-constraint approaches, which use additional auxiliary loss terms describing the constraints. However, catastrophic safety-critical failures might occur in out-of-distribution (OOD) scenarios. This work integrates the flexibility of IL with hard constraint handling in optimal control. Our approach constitutes a general framework for constraint robotic motion planning and control using offline IL. Hard constraints are integrated into the learning problem in a differentiable manner, via explicit completion and gradient-based correction. Simulated experiments of mobile robot navigation and automated driving provide evidence for the performance of the proposed method.
This work proposes a fast iterative method for local steric Poisson--Boltzmann (PB) theories, in which the electrostatic potential is governed by the Poisson's equation and ionic concentrations satisfy equilibrium conditions. To present the method, we focus on a local steric PB theory derived from a lattice-gas model, as an example. The advantages of the proposed method in efficiency are achieved by treating ionic concentrations as scalar implicit functions of the electrostatic potential, though such functions are only numerically achievable. The existence, uniqueness, boundness, and smoothness of such functions are rigorously established. A Newton iteration method with truncation is proposed to solve a nonlinear system discretized from the generalized PB equations. The existence and uniqueness of the solution to the discretized nonlinear system are established by showing that it is a unique minimizer of a constructed convex energy. Thanks to the boundness of ionic concentrations, truncation bounds for the potential are obtained by using the extremum principle. The truncation step in iterations is shown to be energy and error decreasing. To further speed-up computations, we propose a novel precomputing-interpolation strategy, which is applicable to other local steric PB theories and makes the proposed methods for solving steric PB theories as efficient as for solving the classical PB theory. Analysis on the Newton iteration method with truncation shows local quadratic convergence for the proposed numerical methods. Applications to realistic biomolecular solvation systems reveal that counterions with steric hindrance stratify in an order prescribed by the parameter of ionic valence-to-volume ratio. Finally, we remark that the proposed iterative methods for local steric PB theories can be readily incorporated in well-known classical PB solvers.
Graph Neural Networks (GNNs) have recently become increasingly popular due to their ability to learn complex systems of relations or interactions arising in a broad spectrum of problems ranging from biology and particle physics to social networks and recommendation systems. Despite the plethora of different models for deep learning on graphs, few approaches have been proposed thus far for dealing with graphs that present some sort of dynamic nature (e.g. evolving features or connectivity over time). In this paper, we present Temporal Graph Networks (TGNs), a generic, efficient framework for deep learning on dynamic graphs represented as sequences of timed events. Thanks to a novel combination of memory modules and graph-based operators, TGNs are able to significantly outperform previous approaches being at the same time more computationally efficient. We furthermore show that several previous models for learning on dynamic graphs can be cast as specific instances of our framework. We perform a detailed ablation study of different components of our framework and devise the best configuration that achieves state-of-the-art performance on several transductive and inductive prediction tasks for dynamic graphs.
小貼士
登錄享
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191