M. Asada, E. Uchibe, S. Noda, S. Tawaratsumida, and K. Hosoda. Vision-based behavior acquisition for a shooting robot by using a reinforcement learning. In Proc. of IAPR/IEEE Workshop on Visual Behaviors, 1994.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....of the states rather than actions in states. The relative values of actions will be poorly represented, resulting in an unsatisfactory controller. This is compounded as the time intervals between control actions get smaller. The state action deviation problem is closely related to this problem [1]. Advantage Learning [10] addresses the issue of action similarity by emphasising the di#erences in value between the actions. In advantage learning the value of the optimal action is the same as for Q learning, but the lesser value of non optimal actions is emphasised by a scaling factor (k # ....

.... avoidance algorithm was presented in [13] Visual servoing through reinforcement learning was demonstrated in [23] Learned wall avoidance in a constructed environment with discrete actions is described in [22] Asada has developed several vision based reinforcement learning systems, for example [1]. These systems have used discrete actions and adaptively discretised the state space. More recently a visual servoing system using reinforcement learning with continuous states and actions was presented in [21] 7 Conclusion We have demonstrated wandering and servoing behaviours on a real ....

M. Asada, E. Uchibe, S. Noda, S. Tawaratsumida, and K. Hosoda. Vision-based behavior acquisition for a shooting robot by using a reinforcement learning. In Proc. of IAPR/IEEE Workshop on Visual Behaviors, 1994.

....Hosoda Dept. of Mech. Eng. for Computer Controlled Machinery Osaka University, 2 1, Yamadaoka, Suita, Osaka 565, Japan asada robotics.ccm.eng.osaka u.ac.jp Abstract A method is proposed which acquires a purposive behavior of shooting a ball into the goal avoiding collisions with an enemy. In [ Asada et al. 1994], we have presented the soccer robot which learned to shoot a ball into the goal without any enemy, using the Q learning, one of the reinforcement learning methods. Since a simple extension of the method is not practical due to its huge state space, two di erent behaviors each of which are ....

....and less dynamic [ Maes and Brooks, 1990; Connel and Mahadevan, 1993a ] In order to make the role of the reinforcement learning evident in realizing autonomous agents, we need more applications in more dynamic and complex environments. As one of these applications, we built a soccer robot [ Asada et al. 1994 ] that tried to shoot a ball into the goal by applying the Q learning, one of the reinforcement learning schemes which is widely used [Watkins and Dayan, 1992 ] The robot could learn a shooting behavior without world knowledge such as 3 D locations and sizes of the goal and ball in the eld or ....

[Article contains additional citation context not shown here]

M. Asada, S. Noda, S. Tawaratsumida, and K. Hosoda. \Vision-based behavior acquisition for a shooting robot by using a reinforcement learning". In Proc. of IAPR / IEEE Workshop on Visual Behaviors1994, pages 112-118, 1994.

....role from a viewpoint of how to bring up a robo infant. Almost of the all existing methods in robot learning assume the xed environment for their individual tasks and have not considered to change the environments so as to make the animat learn more complicated behaviors step by step. In [10], we proposed the Learning from Easy Missions (hereafter, LEM) paradigm in which the reinforcement learning time can be reduced from the exponential order into the linear order of the state size by placing the animat near the goal state at the beginning and farther from it later. Although the ....

....up an animat, a one eyed mobile robo infant which has visual tracking routines capable of realtime acquisition of optical ow of the environment that are used as visual cues to be associated with several behaviors. We extend the LEM paradigm from only the animat s placement in the xed environment [10] to changing the environments so that the animat can learn complicated behaviors gradually. At the beginning the animat learns the sensorimotor apparatus in its babyhood with almost no obstacles, and then learns several behaviors such as detecting and avoiding obstacles, and pursuit a target in ....

[Article contains additional citation context not shown here]

M. Asada, S. Noda, S. Tawaratsumida, and K. Hosoda. \Vision-Based Behavior Acquisition For A Shooting Robot By Using A Reinforcement Learning". In Proc. of IAPR / IEEE Workshop on Visual Behaviors-1994, pp. 112-118, 1994.

No context found.

Asada, M., S. Noda, S. Tawaratsumida and K. Hosoda (1994b). \Vision-based behavior acquisition for a shooting robot by using a reinforcement learning".

....function with respect to the obstacle tracking behavior among the categories except for C t . 5 Experimental results for a real system 5.1 A con guration of the system Figure 8 shows a con guration of the real mobile robot system. We have constructed the radio control system of the robot [ Asada et al. 1994 ] The image processing and the vehicle control system are operated by VxWorks OS on MVME167(MC68040 CPU) computer which are connected with host Sun workstations via Ether net. The image taken by a TV camera mounted on the robot is transmitted to a UHF receiver and subsampled by the p p i obs ....

M. Asada, S. Noda, S. Tawaratsumida, and K. Hosoda. \Vision-based behavior acquisition for a shooting robot by using a reinforcement learning". In Proc. of IAPR / IEEE Workshop on Visual Behaviors-1994, pages 112-118, 1994.

....real robot applications are reported, which are simple and less dynamic [6, 7] In order to make the role of the reinforcement learning evident in realizing autonomous agents, we need more applications in more dynamic and complex environments. As one of these applications, we built a soccer robot [8] that tried to shoot a ball into the goal by applying the Q learning, a widely used reinforcement learning scheme [9] The robot could learn a shooting behavior without world knowledge such as 3 D locations and sizes of the goal and ball in the eld or the kinematics and dynamics of the robot ....

....which are concurrent with each other but not independent of each other. Such an example is to shoot a ball into a goal avoiding an enemy. The reason why challenging is twofold; ffl from a viewpoint of building a real robot in a real situation, it is more dynamic and complicated environment than in [8], and ffl from a viewpoint of robot learning, existing works have not demonstrated the ability to use previously learned knowledge to speed up the learning of a new policy [10] To the best of our knowledge, only a few works related to the problem have been presented. Whitehead et al. 5] proposed ....

[Article contains additional citation context not shown here]

M. Asada, S. Noda, S. Tawaratsumida, and K. Hosoda. \Vision-based behavior acquisition for a shooting robot by using a reinforcement learning". In Proc. of IAPR / IEEE Workshop on Visual Behaviors-1994, pages 112-118, 1994.

....Boards (Fujitsu Tracking Visioin) MVME 167 MC 68040 RAM P I O (Printer Port) Video Boards MVME 167 Wireless Servo Controller Real robot Fig.3: A con guration of the real system. Fig.3 shows a con guration of the real mobile robot system. We have constructed the radio control system of the robot[14]. The image processing and the vehicle control system are operated by VxWorks OS on MVME167(MC68040 CPU) computer which is connected with host Sun workstations via Ether net. The image taken by a TV camera mounted on the robot is transmitted to a UHF receiver and subsampled by scan line ....

M. Asada, S. Noda, S. Tawaratsumida, and K. Hosoda. \Vision-based behavior acquisition for a shooting robot by using a reinforcement learning". In Proc. of IAPR / IEEE Workshop on Visual Behaviors-1994, pages 112-118, 1994.

....Acquisition On A Real Robot By A Vision Based Reinforcement Learning Minoru Asada, Shoichi Noda, Sukoya Tawaratsumida, and Koh Hosoda Dept. of Mech. Eng. for Computer Controlled Machinery Osaka University, 2 1, Yamadaoka, Suita, Osaka 565, Japan asada robotics.ccm.eng.osaka u. ac.jp Abstract In [1], we have presented the soccer robot which had learned to shoot a ball into the goal using the Q learning. In this paper, we discuss several issues in applying the Qlearning method to a real robot with vision sensor. First, to speed up the learning rate, we implement a mechanism of Learning form ....

....Therefore, it seems di cult to discriminate the both from only the image. Then, they proposed a method to cope with this problem by adopting the internal states and separating action commands into Action frame and Attention frame commands. However, they have not shown the real experiments. In [1], we have presented a soccer robot which had learned to shoot a ball into the goal using the Qlearning. The robot does not need to know any parameters of the 3 D environment or its kinematics dynamics. Information about the changes of the environment is only the image captured from a single TV ....

M. Asada, S. Noda, S. Tawaratsumida, and K. Hosoda. \Vision-based behavior acquisition for a shooting robot by using a reinforcement learning ". In Proc. of IAPR / IEEE Workshop on Visual Behaviors-1994, pages 112-118, 1994.

No context found.

M. Asada, E. Uchibe, S. Noda, S. Tawaratsumida, and K. Hosoda. Vision-based behavior acquisition for a shooting robot by using a reinforcement learning. In Proc. of IAPR/IEEE Workshop on Visual Behaviors, 1994.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC