In this paper we have considered systems in which agents communicate via their environment. In these systems the agents do not have direct and explicit communication with each other and instead they have implicit comm...
详细信息
In this paper we discuss a learning approach to distributed object pushing. In the proposed approach, first the required individual skills for single-robot object pushing are learned using a fuzzy reinforcement learni...
详细信息
In this paper we discuss a learning approach to distributed object pushing. In the proposed approach, first the required individual skills for single-robot object pushing are learned using a fuzzy reinforcement learning method. Then, the robots learn how to coordinate their actions to push the object to the desired configuration cooperatively in a distributed manner. The proposed team-level learning benefits from the knowledge, which is in the form of a Q-table, that the agent has gained in its individual learning phase by a special design of reward signal and state-action representation. Each robot learns a threshold on its Q-value using a single state reinforcement learning method and pushes the object when the Q-value of its best action in the current state is above this threshold. The reward signal is designed based on the robots' Q-tables and no external critic is needed for learning cooperation. Simulation results show that the robots learn their individual skills and a cooperation protocol to push the object cooperatively
In this paper we have considered systems in which agents communicate via their environment. In these systems the agents do not have direct and explicit communication with each other and instead they have implicit comm...
详细信息
In this paper we have considered systems in which agents communicate via their environment. In these systems the agents do not have direct and explicit communication with each other and instead they have implicit communication which is performed via the environment they are located in. Different kinds of environment elements are considered to be changed while on the other hand for all of these situations there is a change in the environment and this change means potential stimuli that fires behaviors of agents for future. It is considered that if we can find a track of changes in the environment and find the roots of a movement in an agent that is originally made by another agent of the group
In this paper, we analyze a dynamic object manipulation (DOM) problem; throwing upward and catching a disc using two planar manipulators. The manipulators control the disc xy position dynamically and use its angular m...
详细信息
In this paper, we analyze a dynamic object manipulation (DOM) problem; throwing upward and catching a disc using two planar manipulators. The manipulators control the disc xy position dynamically and use its angular movement as a free parameter. We use a simple model for the system. Each manipulator has three links and three active joints. In addition, the disc has considerable radius and mass. Therefore, we design a simple mechanism in the palm to damp a portion of the impact
Several path-planning algorithms for mobile robots have been introduced. Proper architectures for mobile robots to implement the path-planning algorithms are also of interest. If the mobile robots are to perform compl...
详细信息
Several path-planning algorithms for mobile robots have been introduced. Proper architectures for mobile robots to implement the path-planning algorithms are also of interest. If the mobile robots are to perform complicated tasks including complex sensing and planning operations and have accepted performance, must be autonomous: capable of acquiring information and performing tasks without programmatic intervention. In this paper we employ a layered architecture for mobile robots to perform our previously introduced cellular automata based path planning technique. It employs an abstraction approach which makes the complexity manageable. The architecture has an important feature which is its internal artifacts; it has some beliefs about the world and these beliefs are represented in artifacts and most actions are planned and performed with respect to these artifacts
Decision trees, being human readable and hierarchically structured, provide a suitable mean to derive state-space abstraction and simplify the inclusion of the available knowledge for a reinforcement learning (RL) age...
详细信息
Decision trees, being human readable and hierarchically structured, provide a suitable mean to derive state-space abstraction and simplify the inclusion of the available knowledge for a reinforcement learning (RL) agent. In this paper, we address two approaches to combine and purify the available knowledge in the abstraction trees, stored among different RL agents in a multi-agent system, or among the decision trees learned by the same agent using different methods. Simulation results in nondeterministic football learning task provide strong evidences for enhancement in convergence rate and policy performance
In this paper, a unique road contour extraction approach from high resolution satellite image is proposed, in which the road contour was extracted in two steps. Firstly, support vector machines (SVM) was employed mere...
详细信息
In this paper, a unique road contour extraction approach from high resolution satellite image is proposed, in which the road contour was extracted in two steps. Firstly, support vector machines (SVM) was employed merely to classify the image into two groups of categories: a road group and a non-road group. The identified road group images are the discrete and irregularly distributed sampled points, and they are an uncompleted data set for the road. Secondly, the road contour was extracted from the road group images using the tensor voting framework, since the tensor voting technique is superior to the traditional methods in extracting the geometrical structure from the uncompleted data set. The experimental results on the high resolution satellite image demonstrate that the proposed approach worked well with images comprised by both rural and urban area features.
Learning distributed object grasp by a group of robots with redundant members is the main focus of this paper. In Elahibakhsh, A. H., et al. (2004), we tackled the problem of learning form closure grasp for planar con...
详细信息
Learning distributed object grasp by a group of robots with redundant members is the main focus of this paper. In Elahibakhsh, A. H., et al. (2004), we tackled the problem of learning form closure grasp for planar convex objects by multiple non-communicating robots without any information about the shape of objects. In this paper, the problem in presence of redundant agents is investigated. Agents' states and actions are designed such that the group learns grasping different objects using Q-learning method. As the environment is not intelligent enough to assess each agent's effect on the team performance, a credit assignment algorithm based on knowledge evaluation is designed. The proposed method considers the environment credit for the team, number of redundant agents, and the expertness level of each agent in its credit assignment. Applicability of the designed approach is verified through a set of simulations. It is shown that the team learns grasping different objects. Therefore, it is expected that the proposed method can be extended for distributed grasp of deformable objects
This paper proposes some methods to improve the fault-tolerance in distributed systems specifically in deterministic situations by distributed decision-making and coordination. Providing a distributed system with faul...
详细信息
This paper proposes some methods to improve the fault-tolerance in distributed systems specifically in deterministic situations by distributed decision-making and coordination. Providing a distributed system with fault tolerance is a feasible but hard problem due to the intrinsic aspects of such systems such as: independency, unpredictability and communication problem. A multi-agent system as an instance of distributed system, can handle different kinds of fault by using traditional fault tolerance techniques. But what focused in this paper are agent-based techniques. In fact proposed methods are based on agent-based help provision by distributed cooperation among helper agents. The helpers try to tune their normal roles such that they can undertake the faulty agents' tasks too. These helpers go through a sub-optimal task selection algorithm, to decide whom to help. It is important to remark that there is no explicit interaction; instead they coordinate their decisions implicitly by adopting the most appropriate task in terms of their speed, relative reliability and the task's criticality coefficient. Proposed ideas are tested on a DCS-like tested to improve its fault tolerance. The results illustrate the effectiveness of the approaches in comparison to the case of no help situation and the case of using purely redundant components.
Despite its potential advantages for fMRI analysis, fuzzy C-means (FCM) clustering suffers from limitations such as the need for a priori knowledge of the number of clusters, and unknown statistical significance and i...
详细信息
Despite its potential advantages for fMRI analysis, fuzzy C-means (FCM) clustering suffers from limitations such as the need for a priori knowledge of the number of clusters, and unknown statistical significance and instability of the results. We propose a randomization-based method to control the false positive rate and estimate statistical significance of the FCM results. Using this novel approach, we develop an fMRI activation detection method. The ability of the method in controlling the false positive rate is shown by analysis of false positives in activation maps of resting-state fMRI data. controlling the false positive rate in FCM allows comparison of different fuzzy clustering methods, using different feature spaces, to other fMRI detection methods. In this paper, using simulation and real fMRI data, we compare a novel feature space that takes the variability of the hemodynamic response function into account (HRF-based feature space) to the conventional cross-correlation analysis and FCM using the cross-correlation feature space.
暂无评论