High cost of environmental interaction and low data efficiency limit the development of reinforcement learning in robotic grasping. This paper proposes an end-to-end robotic grasping method based on offline reinforcem...
High cost of environmental interaction and low data efficiency limit the development of reinforcement learning in robotic grasping. This paper proposes an end-to-end robotic grasping method based on offline reinforcement learning via sequence modeling. It considers the most recent n-step history to assist the agent in making decisions, where a predictive model learns to directly predict actions from raw image inputs. The experimental results show that our method can achieve higher grasping success rate with less training data than traditional reinforcement learning algorithms in offline setting.
The efficient management of storage has become a critical determinant of operational efficiency due to the growing and diverse demand for steel logistics parks. However, the current storage methods consistently result...
The efficient management of storage has become a critical determinant of operational efficiency due to the growing and diverse demand for steel logistics parks. However, the current storage methods consistently result in significant imbalances in the loading of steel products and operating times across multiple yards. To address this challenge, we propose optimizing the distribution of steel products based on demand levels and retrieval efficiency. Firstly, we present an optimization problem that employs demand-driven storage principles to formulate the scheduling of steel product storage. Then, we introduce a novel algorithm called the Adaptive Preferred Evolutionary Heuristic (APEH) to tackle this problem. The results from our experiments indicate that this model effectively enhances the storage structure, and the proposed algorithm consistently yields optimal solutions within a reasonable timeframe.
Steel plates are highly customized according to different customer demands. In this case, the outbound date, the specifications of the steel plates and the distribution of stacks must be simultaneously considered when...
Steel plates are highly customized according to different customer demands. In this case, the outbound date, the specifications of the steel plates and the distribution of stacks must be simultaneously considered when the steel plates are stored in the yard. Improper storage of steel plates significantly increases the number of shuffles during outbound operations. To overcome the challenge, we propose an efficient steel plate storage scheme based on the actual scenario of existing steel plate distribution in the yard. Firstly, a model for steel plate storage scheduling is established to minimize the number of extra blocking plates. Then, three heuristic algorithms are designed to solve the problem. Comparative experiments are carried out to verify the effectiveness of the proposed algorithm. The experimental results show that the proposed method can sufficiently improve the efficiency of steel plates scheduling.
Pickup vehicle scheduling in the steel logistics park is a critical issue in determining the outbound effciency of steel products. Steel products are distributed in the yards of the steel logistics park with mixed sto...
Pickup vehicle scheduling in the steel logistics park is a critical issue in determining the outbound effciency of steel products. Steel products are distributed in the yards of the steel logistics park with mixed storage, and the optional yards for each pickup operation are not unique, which greatly increases the scheduling complexity. To overcome this challenge, in this paper, we propose a pickup vehicle scheduling problem with mixed steel storage (PVSP-MSS) to optimize makespan of pickup vehicle and makespan of steel logistics park simultaneously. The optimization problem is described as a multi-objective problem, and an enhanced Strength Pareto Evolutionary Algorithm 2 (ESPEA) is proposed to solve the problem with high efficiency. Experiments are executed based on data collected from a real steel logistics park. The results confirm that ESPEA can significantly reduce both makespan of each pickup vehicle and makespan of steel logistics park.
Deducing the twisting pattern of spreader is very important for crane operators in steel warehouse, since crane spreaders have to be rotated repeatedly at steel bar transport sites. To improve the efficiency of the st...
Deducing the twisting pattern of spreader is very important for crane operators in steel warehouse, since crane spreaders have to be rotated repeatedly at steel bar transport sites. To improve the efficiency of the steel bar delivery, the state prediction for crane spreaders is necessary. In order to fulfill this requirement, the HTSK-LN-ReLU model prediction method based on Optuna optimization is applied in this paper. We deploy the IMU on the spreader surface, and then obtain the state data during the operation of the crane spreader by wireless transmission. The data are first denoised through fourier filtering, and then the Optuna framework is used to optimize the HTSK-LN-ReLU model's hyperparameters. The evaluation index of mean square error is used for model evaluation, and finally the optimal hyperparameter combination of the model is selected from 200 comparison experiments. Experimental results demonstrate superior performance of the HTSK-LN-ReLU model in predicting the state of crane spreader for steel bar delivery.
In this paper, we revisit the interval observer design problem of harmonic oscillator system with unknown but bounded external disturbance and measurement noise. With the bounding information of the external disturban...
详细信息
Reinforcement Learning (RL) has been applied to robotic arm control, which enables the agent to learn an effective policy to solve complex tasks. However, it requires constant interaction with the environment leading ...
Reinforcement Learning (RL) has been applied to robotic arm control, which enables the agent to learn an effective policy to solve complex tasks. However, it requires constant interaction with the environment leading to low sample efficiency. In this paper, we propose a robotic arm control approach based on planning via lookahead search, which is a model-based RL algorithm to improve the sample efficiency. The approach builds an environment model in order to obtain the dynamics of the environment. Thus the model can be used to plan future actions by a tree-based search. The experiments show that our approach can solve the task of robotic arm control with less environmental samples.
A new ball joint actuator with three-degree-of-freedom is proposed to solve the problem of space motion. It can perform some specific work in a certain space with multiple degrees of freedom. At first, the kinematics ...
详细信息
This paper investigates the problem of designing control policies that satisfy high-level specifications described by signal temporal logic (STL) in unknown, stochastic environments. While many existing works concentr...
详细信息
ISBN:
(数字)9798350384574
ISBN:
(纸本)9798350384581
This paper investigates the problem of designing control policies that satisfy high-level specifications described by signal temporal logic (STL) in unknown, stochastic environments. While many existing works concentrate on optimizing the spatial robustness of a system, our work takes a step further by also considering temporal robustness as a critical metric to quantify the tolerance of time uncertainty in STL. To this end, we formulate two relevant control objectives to enhance the temporal robustness of the synthesized policies. The first objective is to maximize the probability of being temporally robust for a given threshold. The second objective is to maximize the worst-case spatial robustness value within a bounded time shift. We use reinforcement learning to solve both control synthesis problems for unknown systems. Specifically, we approximate both control objectives in a way that enables us to apply the standard Q-learning algorithm. Theoretical bounds in terms of the approximations are also derived. We present case studies to demonstrate the feasibility of our approach.
Wood broken defects, such as cracks, penetrating dead knots and wanes, will seriously weaken the strength of the wood and destroy the structral integrity of the wood. Therefore, these defects should be inspected and t...
详细信息
Wood broken defects, such as cracks, penetrating dead knots and wanes, will seriously weaken the strength of the wood and destroy the structral integrity of the wood. Therefore, these defects should be inspected and then cut during the production. Although vision-based detection methods have been proposed to detect wood defect regions, they can not detect the defects that change in depth. To solve this problem, this paper focuses on the defect detection by measuring depth difference through 3D laser data. First, the 3D laser data is preprocessed to remove the noise. Secondly, we construct a support profile by multiple mean filtering at the line-level. Next, the residual components are obtained by subtracting the support profile from the profile data and the defect is determined by comparing the residual components and the setting threshold. Finally, we perform merging and filtering operation to determine the final position of defects. The experimental results verify the effectiveness of the proposed method.
暂无评论