Flat objects with negligible thicknesses like books and disks are challenging to be grasped by the robot because of the width limit of the robot's gripper, especially when they are in cluttered environments. Pre-g...
Flat objects with negligible thicknesses like books and disks are challenging to be grasped by the robot because of the width limit of the robot's gripper, especially when they are in cluttered environments. Pre-grasp manipulation is conducive to rearranging objects on the table and moving the flat objects to the table edge, making them graspable. In this paper, we formulate this task as Parameterized Action Markov Decision Process, and a novel method based on deep reinforcement learning is proposed to address this problem by introducing sliding primitives as actions. A weight-sharing policy network is utilized to predict the sliding primitive's parameters for each object, and a Q-network is adopted to select the acted object among all the candidates on the table. Meanwhile, via integrating a curriculum learning scheme, our method can be scaled to cluttered environments with more objects. In both simulation and real-world experiments, our method surpasses the existing methods and achieves pre-grasp manipulation with higher task success rates and fewer action steps. Without fine-tuning, it can be generalized to novel shapes and household objects with more than 85% success rates in the real world. Videos and supplementary materials are available at https://***/view/pre-grasp-sliding.
This paper presented a p-norm constraint Lorentzian adaptive algorithm to estimate the sparse multi-path channel ina-stable noise environment. We derived the proposed algorithm based on the Lagrange multiplier method ...
详细信息
This paper addresses the problem of maneuvering multi-target tracking by a network of sensors having different and limited fields of view (FoV s). Each local sensor runs the Gaussian Mixture Probability Hypothetical D...
详细信息
ISBN:
(纸本)9781665468893
This paper addresses the problem of maneuvering multi-target tracking by a network of sensors having different and limited fields of view (FoV s). Each local sensor runs the Gaussian Mixture Probability Hypothetical Density (GMPHD) filter. Due to the target maneuver, based on a single motion model, severe tracking performance degradation can be observed. We propose to use multi-model (MM) method to realize the adaptation to motion characteristic so as to overcome maneuverability of targets. Then considering FoV s of sensors in the network are different and limited, the standard weighted arithmetic average (WAA) fusion is no longer applicable and leads to an underestimation of the target number. Therefore, we use state-dependent WAA (SD-WAA) fusion rule, which performs the WAA in a more robust way by calculating a set of state-related fusion weights. Numerical experiment is designed to demonstrate the efficacy of the proposed method.
Event-triggered control has attracted considerable attention for its effectiveness in resource-restricted applications. To make event-triggered control as an end-to-end solution, a key issue is how to effectively lear...
详细信息
A decoupled control approach based on linear active disturbance rejection control idea is proposed. The plant model, in which the coupling dynamics is as a part of disturbance, is established, and some virtual control...
详细信息
The monitoring and fault detection for the Czochralski (Cz) crystal growth process is important to ensure the quality of the produced mono-crystalline silicon. The Cz process is a typical multi-modes industrial proces...
详细信息
Aiming at the fusion problem of multi-source heterogeneous dynamic data sources, based on the subjective and objective comprehensive weighting idea, an efficient multi-source dynamic data fusion algorithm is proposed....
Aiming at the fusion problem of multi-source heterogeneous dynamic data sources, based on the subjective and objective comprehensive weighting idea, an efficient multi-source dynamic data fusion algorithm is proposed. First, the outlier detection method based on linear regression model is used to preprocess the multi-source dynamic time series data. Then, determine the weight of the dynamic data. The prior weight of the data source is determined by the AHP method, cyclic scoring method, etc., and the posterior weight is determined according to the similarity between the time series data. Based on the Euclidean distance, a similarity measure between different data sources and a similarity measure between different dynamic data samples are constructed respectively. The former is used for the fusion of dynamic data standard deviation expectations, and the latter is used for the fusion of mean expectations. Finally, the comprehensive weight of each dynamic time series of each data source is obtained, and the comprehensive weighted fusion method is used to obtain the expected value and standard deviation of the dynamic response. The analysis of the calculation example shows that the proposed algorithm has higher computational efficiency than the method that treats dynamic data as several discrete static data.
Point cloud preprocessing is still a challenging task in the marine environment, for it is difficult to filter out non-obstacle points while avoiding damage to the obstacle completeness. In this paper, we propose a no...
详细信息
To address the failure of precise overload tracking and anti-interference caused by the difficulty of accurate modeling of a complex aircraft, the controller designing method based on deep reinforcement learning is st...
详细信息
ISBN:
(数字)9781728159225
ISBN:
(纸本)9781728159232
To address the failure of precise overload tracking and anti-interference caused by the difficulty of accurate modeling of a complex aircraft, the controller designing method based on deep reinforcement learning is studied. This paper trained the control network based on the Proximal Policy Optimization (PPO), studied the tracking control problem of the aircraft, and accurately tracked the typical command signals. Fixed-point simulation of the aircraft is performed, with results showing that, in presence of aircraft model parameter variation and external disturbance, the controller based on deep reinforcement learning can achieve accurate tracking of overload commands.
Air combat decision-making is a critical issue in Unmanned Air Vehicle automatic combat. Precise and efficient maneuvering strategies are extremely important for the final victory. Quantitative research has been condu...
详细信息
ISBN:
(数字)9781728180250
ISBN:
(纸本)9781728180267
Air combat decision-making is a critical issue in Unmanned Air Vehicle automatic combat. Precise and efficient maneuvering strategies are extremely important for the final victory. Quantitative research has been conducted on the maneuver strategy. In this study, a knowledge-based maneuver action library for air combat was established by the Rough Set Theory, which can help quick response to the battlefield situation. Since not all influence factors of battlefield situation are that significant and the computing source is finite, the rough set model was simplified to increase reaction rate. This paper introduced an advanced genetic algorithm to reduce the attributes of rough sets, taking the purity of each attribute into account, so as to make the condition attribute as close to 1 or 0 as possible. As a result, the reducing accuracy of the whole decision-making system is improved. The simulation results show that the rough set model built in this paper is efficient, feasible and reasonable, and the algorithm can support the air combat maneuver strategy decision system.
暂无评论