R. Salomon. The evolution of different neuronal control structures fo autonomous agents. Robotics and Autonomous Systems, (in press). Home/Search Document Details and Download
Summary Related Articles Check |
This paper is cited in the following contexts:
Learning from innate behaviors: A Quantitative Evaluation of.. - Sharkey (1998) (4 citations) (Correct)
....here assess the development of the use of a prewired or innate controllers to bootstrap learning in a Multi Layer Perceptron to navigate a mobile robot to goals in an obstacle laden environment. Goal finding while avoiding obstacles is a common behavioral benchmark for robot controllers, e.g. [1, 7, 22, 11, 33, 26, 23] and it has been employed in autonomous robotics for several decades, e.g. Gray Walter [14] While it seems clear that neural networks could be trained to perform the task, there are two novel questions here. First, could an improvement be gained over the performance of simple innate hardwired ....
....Some quantitative measures have already been used to assess or determine the course of learning avoidance and goalfinding. In evolutionary learning, fitness has been determined by parameters such as speed, straight direction, and obstacle avoidance [11] or speed and maximum sensor distance [26]. In reinforcement learning, Meeden [22] measured the amount of punishment received by different controllers and used this to assess the progress of learning and the difference between two controllers. Touzet [33] used two main measures to measure the performance of the controllers: distance to ....
R. Salomon. The evolution of different neuronal control structures fo autonomous agents. Robotics and Autonomous Systems, (in press).
A Multi-agent based Evolutionary Artificial Neural Network.. - Wang, Mckenzie (1999) (Correct)
.... projects use Genetic Algorithms (GAs) to evolve neural control structures [2, 3] Nonetheless, recent studies have disclosed that GA is very time consuming when the parameters to be optimized exhibit epistasis a nonlinear interaction of parameters with respect to the fitness of an individual [9, 10]. Subsequent results have also shown that another evolutionary algorithm, Evolution Strategies (ES) may be an alternative to overcome this problem. ES is especially designed for applications that involve real valued parameters [8, 11] A key concept of ES is that both control parameters x and ....
R. Salomon. The Evolution of Different Neuronal Control Structures for Autonomous Agents. Robotics and Autonomous Systems, Vol.22, No.3-4, pp.199-213, 1997.
Radial Basis Function Networks for Autonomous Agent Control - Salomon (1997) Self-citation (Salomon) (Correct)
....it does not exhibit overlearning due to a too long learning process. In addition, the next section shows that this model has an incremental learning behavior in the sense that the presentation of patterns from a subspace does not affect the mapping of other patterns. 4 Results Previous research [13], in which evolutionary algorithms evolve different neural controllers, has shown that RBF networks are well suited for autonomous agents. Preliminary experiments with the proposed model indicate that it yields a high degree of self organization. Since we are currently developing an appropriate ....
R. Salomon, The Evolution of Different Neuronal Control Structures for Autonomous Agents, Accepted for publication in N. Sharkey (Ed.), Robotics and Autonomous Systems; Robot Learning: The New Wave (Special Issue), Elsevier.
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