The thesis is concerned with modeling and simulation of autonomous robots in order to investigate emergent behavior of situated agents interacting with their environment. First, models are proposed to capture the essentials of real-world physics for simple objects and their interaction within the world. These models are then embedded into a coherent framework (DEVS) to allow their discrete-event driven execution on a computer. In the remaining part, the notion of homeokinesis is discussed in detail and proves to provide a sound framework for emergence of behavior. Experiments are conducted at this stage, which utilize the simulator developed in the first part. Resulting emergent behavior is presented and investigated in detail. Finally an encountered limitation in the orginal formulation of homeokinetic control is remedied by application of recurrent neural networks.
Simulation and investigation of embodied autonomous robots driven by homeokinetic control
2001-04-01
Theses
Electronic Resource
English
Collective control of modular soft robots via embodied Spiking Neural Cellular Automata
BASE | 2022
|