Papers by robert griffioen
2008 Ieee Congress on Evolutionary Computation, Jun 1, 2008
In this paper we present a new controller architecture. A central design choice is that the contr... more In this paper we present a new controller architecture. A central design choice is that the controller can be easily modified or changed by evolution. Our aim is to initially endow the agents with as little knowledge as possible and to let them evolve their controllers autonomously. One particular aspect of the controller that we will investigate in this research is the evolution of state persistent controllers. With this is meant controllers that that can carry out multiple tasks. Without state persistency, agents may suffer from socalled "unfocused attention": the case where an agent is caught in the middle between tasks and interchangeably executes these partially, but can and will never fully commit to either one and therefore never accomplish any. We will present the statepersistent controller architecture and demonstrate this property in an experiment.
Lecture Notes in Computer Science, 2003
We introduce selfrepairing neural networks as a model for recovery from brain damage. Small lesio... more We introduce selfrepairing neural networks as a model for recovery from brain damage. Small lesions are repaired through reinstatement of the redundancy in the network's connections. With mild lesions, this process can model autonomous recovery. Moderate lesions require patterned input. In this paper, we discuss implementations in three types of network of increasing biological plausibility. We also mention some results from random graph theory. Finally, we discuss the implications for rehabilitation theory.
In this paper we present a new controller architecture. A central design choice is that the contr... more In this paper we present a new controller architecture. A central design choice is that the controller can be easily modified or changed by evolution. Our aim is to initially endow the agents with as little knowledge as possible and to let them evolve their controllers autonomously. One particular aspect of the controller that we will investigate in this research is the evolution of state persistent controllers. With this is meant controllers that that can carry out multiple tasks. Without state persistency, agents may suffer from socalled "unfocused attention": the case where an agent is caught in the middle between tasks and interchangeably executes these partially, but can and will never fully commit to either one and therefore never accomplish any. We will present the statepersistent controller architecture and demonstrate this property in an experiment.
* Abstract The NewTies project is implementing a simulation in which societies of agents are expe... more * Abstract The NewTies project is implementing a simulation in which societies of agents are expected to develop autonomously as a result of individual, population and social learning. These societies are expected to be able to solve environmental challenges by acting collectively. The challenges are intended to be analogous to those faced by early, simple, small-scale human societies.
2008 IEEE Congress on Evolutionary Computation (IEEE World Congress on Computational Intelligence), 2008
In this paper we investigate the effects of individual learning on an evolving population of situ... more In this paper we investigate the effects of individual learning on an evolving population of situated agents. We work with a novel type of system where agents can decide autonomously (by their controllers) if/when they reproduce and the bias in the agent controllers for the mating action is adaptable. Our experiments show that in such a system reinforcement learning with the straightforward rewards system based on energy makes the agents loose their interest in mating, that is, learning counteracts evolution. This effect can be eliminated by introducing a specific reward for the mating action that always gives positive feedback to the agents, as some kind of pleasure or "orgasm". Using such a combination, individual learning is able to keep non-optimal agents alive, where evolution only leads to extinction. Despite that it preserves a viable population that is able to acquire the necessary survival skills, we also found a disadvantage of learning, namely a hiding effect of ill adapted non-optimal performing agents.
Proceedings of GECCO 2007: Genetic and Evolutionary Computation Conference, 2007
1 New and Emerging World models Through Individual, Evolutionary and Social learning, EU FP6 Proj... more 1 New and Emerging World models Through Individual, Evolutionary and Social learning, EU FP6 Project, http://www.newties.org 2 For more details see
Journal of Artificial Societies and Social Simulation, Mar 31, 2006
The NewTies project is implementing a simulation in which societies of agents are expected to de-... more The NewTies project is implementing a simulation in which societies of agents are expected to de-velop autonomously as a result of individual, population and social learning. These societies are expected to be able to solve environmental challenges by acting collectively. The challenges are in-tended to be analogous to those faced by early, simple, small-scale human societies. This report on work in progress outlines the major features of the system as it is currently conceived within the project, including the design of the agents, the environment, the mechanism for the evolution of language and the peer-to-peer infrastructure on which the simulation runs.
Uploads
Papers by robert griffioen