Search results

This paper aims to present a new learning from demonstration-based trajectory planner that generalizes and extracts relevant features of the desired motion for an industrial robot.

The proposed trajectory planner is based on the concept of human arm motion imitation by the robot end-effector. The teleoperation-based real-time control architecture is used for direct and effective imitation learning. Using this architecture, a self-sufficient trajectory planner is designed which has inbuilt mapping strategy and direct learning ability. The proposed approach is also compared with the conventional robot programming approach.

The developed planner was implemented on the 5 degrees-of-freedom industrial robot SCORBOT ER-4u for an object manipulation task. The experimental results revealed that despite morphological differences, the robot imitated the demonstrated trajectory with more than 90 per cent geometric similarity and 60 per cent of the demonstrations were successfully learned by the robot with good positioning accuracy. The proposed planner shows an upper hand over the existing approach in robustness and operational ease.

The approach assumes that the human demonstrator has the requisite expertise of the task demonstration and robot teleoperation. Moreover, the kinematic capabilities and the workspace conditions of the robot are known a priori.

The real-time implementation of the proposed methodology is possible and can be successfully used for industrial automation with very little knowledge of robot programming. The proposed approach reduces the complexities involved in robot programming by direct learning of the task from the demonstration given by the teacher.

This paper discusses a new framework blended with teleoperation and kinematic considerations of the Cartesian space, as well joint space of human and industrial robot and optimization for the robot programming by demonstration.

This paper aims to introduce an imitation learning framework for a wheeled mobile manipulator based on dynamical movement primitives (DMPs). A novel mobile manipulator with the capability to learn from demonstration is introduced. Then, this study explains the whole process for a wheeled mobile manipulator to learn a demonstrated task and generalize to new situations. Two visual tracking controllers are designed for recording human demonstrations and monitoring robot operations. The study clarifies how human demonstrations can be learned and generalized to new situations by a wheel mobile manipulator.

The kinematic model of a mobile manipulator is analyzed. An RGB-D camera is applied to record the demonstration trajectories and observe robot operations. To avoid human demonstration behaviors going out of sight of the camera, a visual tracking controller is designed based on the kinematic model of the mobile manipulator. The demonstration trajectories are then represented by DMPs and learned by the mobile manipulator with corresponding models. Another tracking controller is designed based on the kinematic model of the mobile manipulator to monitor and modify the robot operations.

To verify the effectiveness of the imitation learning framework, several daily tasks are demonstrated and learned by the mobile manipulator. The results indicate that the presented approach shows good performance for a wheeled mobile manipulator to learn tasks through human demonstrations. The only thing a robot-user needs to do is to provide demonstrations, which highly facilitates the application of mobile manipulators.

The research fulfills the need for a wheeled mobile manipulator to learn tasks via demonstrations instead of manual planning. Similar approaches can be applied to mobile manipulators with different architecture.

This paper aims to study the agent learning from expert demonstration data while incorporating reinforcement learning (RL), which enables the agent to break through the limitations of expert demonstration data and reduces the dimensionality of the agent’s exploration space to speed up the training convergence rate.

Firstly, the decay weight function is set in the objective function of the agent’s training to combine both types of methods, and both RL and imitation learning (IL) are considered to guide the agent's behavior when updating the policy. Second, this study designs a coupling utilization method between the demonstration trajectory and the training experience, so that samples from both aspects can be combined during the agent’s learning process, and the utilization rate of the data and the agent’s learning speed can be improved.

The method is superior to other algorithms in terms of convergence speed and decision stability, avoiding training from scratch for reward values, and breaking through the restrictions brought by demonstration data.

The agent can adapt to dynamic scenes through exploration and trial-and-error mechanisms based on the experience of demonstrating trajectories. The demonstration data set used in IL and the experience samples obtained in the process of RL are coupled and used to improve the data utilization efficiency and the generalization ability of the agent.

This purpose of this paper is to provide an overview of the theoretical background and applications of inverse reinforcement learning (IRL).

Reinforcement learning (RL) techniques provide a powerful solution for sequential decision making problems under uncertainty. RL uses an agent equipped with a reward function to find a policy through interactions with a dynamic environment. However, one major assumption of existing RL algorithms is that reward function, the most succinct representation of the designer's intention, needs to be provided beforehand. In practice, the reward function can be very hard to specify and exhaustive to tune for large and complex problems, and this inspires the development of IRL, an extension of RL, which directly tackles this problem by learning the reward function through expert demonstrations. In this paper, the original IRL algorithms and its close variants, as well as their recent advances are reviewed and compared.

This paper can serve as an introduction guide of fundamental theory and developments, as well as the applications of IRL.

This paper surveys the theories and applications of IRL, which is the latest development of RL and has not been done so far.

In this paper, the authors aim to propose a method for learning robotic assembly sequences, where precedence constraints and object relative size and location constraints can be learned by demonstration and autonomous robot exploration.

To successfully plan the operations involved in assembly tasks, the planner needs to know the constraints of the desired task. In this paper, the authors propose a methodology for learning such constraints by demonstration and autonomous exploration. The learning of precedence constraints and object relative size and location constraints, which are needed to construct a planner for automated assembly, were investigated. In the developed system, the learning of symbolic constraints is integrated with low-level control algorithms, which is essential to enable active robot learning.

The authors demonstrated that the proposed reasoning algorithms can be used to learn previously unknown assembly constraints that are needed to implement a planner for automated assembly. Cranfield benchmark, which is a standardized benchmark for testing algorithms for robot assembly, was used to evaluate the proposed approaches. The authors evaluated the learning performance both in simulation and on a real robot.

The authors' approach reduces the amount of programming that is needed to set up new assembly cells and consequently the overall set up time when new products are introduced into the workcell.

In this paper, the authors propose a new approach for learning assembly constraints based on programming by demonstration and active robot exploration to reduce the computational complexity of the underlying search problems. The authors developed algorithms for success/failure detection of assembly operations based on the comparison of expected signals (forces and torques, positions and orientations of the assembly parts) with the actual signals sensed by a robot. In this manner, all precedence and object size and location constraints can be learned, thereby providing the necessary input for the optimal planning of the entire assembly process.

Under the development trend of intelligent manufacturing, the unstructured environment requires the robot to have a good generalization performance to adapt to the scene changes. The purpose of this paper aims to present a learning from demonstration (LfD) method (task parameterized [TP]-dynamic movement primitives [DMP]-GMR) that combines DMPs and TP-LfD to improve generalization performance and solve object manipulation tasks.

The dynamic time warping algorithm is applied to processing demonstration data to obtain a more standard learning model in the proposed method. The DMPs are used to model the basic trajectory learning model. The Gaussian mixture model is introduced to learn the force term of DMPs and solve the problem of learning from multiple demonstration trajectories. The robot can learn more local geometric features and generalize the learned model to unknown situations by adding task parameters.

An evaluation criterion based on curve similarity calculated by the Frechet distance was constructed to evaluate the model’s interpolation and extrapolation performance. The model’s generalization performance was assessed on 2D virtual data sets, and first, the results show that the proposed method has better interpolation and extrapolation performance than other methods.

The proposed model was applied to the axle-hole assembly task on real robots, and the robot’s posture in grasping and placing the axle part was taken as the task parameter of the model. The experiment results show that The proposed model is competitive with other models.

Teaching by demonstration (TbD) is a promising way for robot learning skills in human and robot collaborative hybrid manufacturing lines. Traditionally, TbD systems have only concentrated on how to enable robots to learn movement skills from humans. This paper aims to develop an extended TbD system which can also enable learning stiffness regulation strategies from humans.

Here, the authors propose an extended dynamical motor primitives (DMP) framework to achieve this goal. In addition to the advantages of the traditional ones, the authors’ framework can enable robots to simultaneously learn stiffness and the movement from human demonstrations. Additionally, Gaussian mixture model (GMM) is used to capture the features of movement and of stiffness from multiple demonstrations of the same skill. Human limb surface electromyography (sEMG) signals are estimated to obtain the reference stiffness profiles.

The authors have experimentally demonstrated the effectiveness of the proposed framework. It shows that the authors approach could allow the robot to execute tasks in a variable impedance control mode with the learned movement trajectories and stiffness profiles.

In robot skill acquisition, DMP is widely used to encode robotic behaviors. So far, however, these DMP modes do not provide the ability to properly represent and generalize stiffness profiles. The authors argue that both movement trajectories and stiffness profiles should be considered equally in robot skill learning. The authors’ approach has great potential of applications in the future hybrid manufacturing lines.

This paper aims to deal with the trade-off of the stability and the accuracy in learning human control strategy from demonstrations. With the stability conditions and the estimated stability region, this paper aims to conveniently get rid of the unstable controller or controller with relatively small stability region. With this evaluation, the learning human strategy controller becomes much more robust to perturbations.

In this paper, the criterion to verify the stability and a method to estimate the domain of attraction are provided for the learning controllers trained with support vector machines (SVMs). Conditions are formulated based on the discrete-time system Lyapunov theory to ensure that a closed-form of the learning control system is strongly stable under perturbations (SSUP). Then a Chebychev point based approach is proposed to estimate its domain of attraction.

Some of such learning controllers have been implemented in the vertical balance control of a dynamically stable, statically unstable wheel mobile robot.

Compared with traditional methods relying on manual teaching or system modeling, data-driven learning methods, such as deep reinforcement learning and imitation learning, show more promising potential to cope with the challenges brought by increasingly complex tasks and environments, which have become the hot research topic in the field of robot skill learning. However, the contradiction between the difficulty of collecting robot–environment interaction data and the low data efficiency causes all these methods to face a serious data dilemma, which has become one of the key issues restricting their development. Therefore, this paper aims to comprehensively sort out and analyze the cause and solutions for the data dilemma in robot skill learning.

First, this review analyzes the causes of the data dilemma based on the classification and comparison of data-driven methods for robot skill learning; Then, the existing methods used to solve the data dilemma are introduced in detail. Finally, this review discusses the remaining open challenges and promising research topics for solving the data dilemma in the future.

This review shows that simulation–reality combination, state representation learning and knowledge sharing are crucial for overcoming the data dilemma of robot skill learning.

To the best of the authors’ knowledge, there are no surveys that systematically and comprehensively sort out and analyze the data dilemma in robot skill learning in the existing literature. It is hoped that this review can be helpful to better address the data dilemma in robot skill learning in the future.

Learning from demonstration (LfD) provides an intuitive way for non-expert persons to teach robots new skills. However, the learned motion is typically fixed for a given scenario, which brings serious adaptiveness problem for robots operating in the unstructured environment, such as avoiding an obstacle which is not presented during original demonstrations. Therefore, the robot should be able to learn and execute new behaviors to accommodate the changing environment. To achieve this goal, this paper aims to propose an improved LfD method which is enhanced by an adaptive motion planning technique.

The LfD is based on GMM/GMR method, which can transform original off-line demonstrations into a compressed probabilistic model and recover robot motion based on the distributions. The central idea of this paper is to reshape the probabilistic model according to on-line observation, which is realized by the process of re-sampling, data partition, data reorganization and motion re-planning. The re-planned motions are not unique. A criterion is proposed to evaluate the fitness of each motion and optimize among the candidates.

The proposed method is implemented in a robotic rope disentangling task. The results show that the robot is able to complete its task while avoiding randomly distributed obstacles and thereby verify the effectiveness of the proposed method. The main contributions of the proposed method are avoiding unforeseen obstacles in the unstructured environment and maintaining crucial aspects of the motion which guarantee to accomplish a skill/task successfully.

Traditional methods are intrinsically based on motion planning technique and treat the off-line training data as a priori probability. The paper proposes a novel data-driven solution to achieve motion planning for LfD. When the environment changes, the off-line training data are revised according to external constraints and reorganized to generate new motion. Compared to traditional methods, the novel data-driven solution is concise and efficient.