Developmental Systems: Papers from the AAAI Fall Symposium
Sanjeev Kumar, Gregory S. Hornby, and Joshua Bongard, Cochairs
In nature, the processes of biological development have been pivotal in nature's ability to construct adaptable, modularized, and self-repairing systems of incredible complexity. The development of multicellular organisms from a single cell provides a plentiful and rich source of knowledge and inspiration for constructing developmental systems that model biological processes or enhance evolutionary design systems. Development biology-inspired approaches represent a method for facilitating the construction of robust, complex adaptive systems in a more modular and evolvable manner than conventional methods. For example the state of the art in evolutionary robotics involves evolving controllers for robots with fixed morphologies, rather than all aspects of the robot. Artificial developmental systems may be useful for overcoming this limitation: biological development relies on coupled growth of all of an organism's subsystems in parallel, and the evolution of modular gene subnetworks and phenotypic modules. This symposium discussed how best to extract the key principles of biological development as they relate to design of complex artifacts and computation in general.