Machines' "many little lives" *: An experimental robotic platform to investigate the fundamental and necessary ingredients for life-like artificial sy

点击翻译按钮获取中文摘要

This project aims to build a soft, three-level artificial platform to systematically test life-like properties like equilibrium, emergence, robustness, and learning. The objectives are listed below and detailed in later sections: 1. Build tensegrity modules with multiple equilibria that will (passively) sense and react to a simple environment2. Implement CPG-like oscillators for modules, capable to sense sensory information across the network and perform short-range communication with adjacent modules to achieve locomotion 3. Connect levels 1 and 2 with artificial neural long-distance connections stimulating the lower levels and investigate whether the system can be trained to express simple forms of learning (e.g., habituation)4. Successful learning would enable us to address how information moves across scales in a system and offer opportunities to investigate the possibility of (i) "artificial chimaeras", i.e., merging two tensegrity modules trained to reach different equilibrium points and observe what the result is, and/or (ii) "robot cancer", that is, how a single module breaking communication with others (as suggested by [16]) will affect the network.