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
作者：
ERIC L. SEVIGNYJONATHAN P. CAULKINSEric L. Sevigny
is a Ph.D. student in the Graduate School of Public and International Affairs at the University of Pittsburgh. Mr. Sevigny's research interests lie in the area of drug policy with an emphasis on enforcement sanctioning international control and treatment/harm reduction. His previous experience includes substance abuse counseling and substance abuse treatment needs research of special populations including prisoners the homeless and adolescents. Mr. Sevigny received a B.A. in Psychology from Middlebury College. Jonathan P. Caulkins
Ph.D. is Professor of Operations Research and Public Policy at Carnegie Mellon University's Heinz School of Public Policy. Dr. Caulkins specializes in mathematical modeling and systems analysis of social policy problems with a particular focus on issues pertaining to drugs crime violence and prevention. Dr. Caulkins has also published on airline operations sulfur dioxide pollution trading markets Internet-based advertising flexible manufacturing systems and personnel performance evaluation among other topics. At RAND he has been a consultant visiting scientist codirector of RAND's Drug Policy Research Center (1994–1996) and founding director of RAND's Pittsburgh office (1999–2001). Dr. Caulkins received a B.S. and M.S. in Systems Science from Washington University and an S.M. in Electrical Engineering and Computer Science and Ph.D. in Operations Research both from M.I.T.
Research Summary: Drug policy reformers and defenders contest the extent to which low-level drug offenders are being sent to prison and for how long. Using data from the Survey of Inmates in Federal and State Correcti...
详细信息
Research Summary: Drug policy reformers and defenders contest the extent to which low-level drug offenders are being sent to prison and for how long. Using data from the Survey of Inmates in Federal and State Correctional Facilities, 1997 (BJS, 2000), we assess the seriousness of incarcerated drug offenders along dimensions of dangerousness, culpability, and harm—specifically, functional role and drug group participation, type and amount of drugs, firearms involvement, and criminal conviction and arrest history. We find that only about 1.6% of federal and 5.7% of state inmates can be described as “unambiguously low-level.” Alternatively, not many are “kingpins.” Rather, most fall into a middle spectrum representing different degrees of seriousness that depend on what factors are emphasized. Policy Implications: Our findings dampen hopes of dramatically reducing prison populations by getting out of prison those who are unambiguously low-level drug offenders. They simply do not represent the majority of incarcerated drug offenders. In particular, most played some role in distribution, so eliminating prison terms for users (decriminalization) would not have affected many now in prison. Indeed, if decriminalization increased demand, it could plausibly increase prison populations by increasing the number of suppliers still subject to imprisonment. On the other hand, “drug courier exceptions” to sentencing laws that apply to minor role offenders possessing large quantities could have a greater prison reduction impact.
This paper proposes an Adaptive Critic Based Neuro-Fuzzy controller (ACBNFC) to Thyristor controlled Series Capacitor (TCSC), which might have a significant impact on power system dynamics. The function of the ACBNFC ...
详细信息
Robust asymptotic stability for hybrid systems is considered. For this purpose, a generalized solution concept is developed. The first step is to characterize a hybrid time domain that permits an efficient description...
详细信息
The rapid growth of the Internet and increased desmand to use the Internet for voice and video applications necessitate the design and utilization of new Internet architectures with effective congestion control algori...
详细信息
This paper provides a model predictive approach to control switched reluctance motors (SRM's). A local linear neuro-fuzzy model is used to model SRM. Then a predictive control schema is devised considering an appr...
详细信息
In this paper a model reference variable structure controller (VSC) for an active suspension system is designed. A half vehicle model is used in which, the vertical and pitch motions of the mass supported by the suspe...
详细信息
Robust asymptotic stability for hybrid systems is considered. For this purpose, a generalized solution concept is developed. The first step is to characterize a hybrid time domain that permits an efficient description...
详细信息
Robust asymptotic stability for hybrid systems is considered. For this purpose, a generalized solution concept is developed. The first step is to characterize a hybrid time domain that permits an efficient description of the convergence of a sequence of solutions. Graph convergence is used. Then a generalized solution definition is given that leads to continuity with respect to initial conditions and perturbations of the system data. This property enables new results on necessary conditions for asymptotic stability in hybrid systems.
For comments by R.J. Mantz and H. De Battisa see ibid. (vol. 51, p. 736-38, June 2004) . For original paper by A. S. Hodel and C. E. Hall see ibid.(vol. 48, p. 442-51, Apr. 2001).
For comments by R.J. Mantz and H. De Battisa see ibid. (vol. 51, p. 736-38, June 2004) . For original paper by A. S. Hodel and C. E. Hall see ibid.(vol. 48, p. 442-51, Apr. 2001).
This paper presents an application of wavelet networks in identification and control design tor a class of non-linear dynamical systems. The technique of feedback linearization, supervisory control and H∞ control are...
详细信息
暂无评论