| Motivated by the autonomous development process of humans, we are interested in building a robot that learns online in real-time by interacting with the environment through sensors and effectors, and develops a representation of the environment and tasks autonomously. We call such a robot a developmental robot, which should satisfy eight challenging requirements: environment openness, high-dimensional sensors, completeness in using sensory information, online processing and real-time speed, incremental processing, performing while learning, and handling increasingly more complex tasks.; To partially satisfy the above-mentioned challenging requirements for a developmental robot, this research contributes: (1) Building the body of a humanoid robot, called Dav. Embodiment is the most overlooked aspect of human intelligence in traditional artificial intelligence. The morphology of the robot is preferably of a humanoid form. This human-like shape may allow humans to interact with the robot more naturally and provide the similar physical constraints as humans during those interactions. Furthermore, directly situating the robot in the physical world relieves the needs for a symbolic intermediary representation. The actual physical world serves for this representation. Therefore, as a part of this research, we built the Dav robot, which is the only untethered mobile humanoid currently in the universities of the United States. (2) Creating a developmental architecture and a modified version of incremental hierarchical discriminant analysis (IHDR) algorithm. This thesis also presents a theory of developmental mental architecture. Five architecture types, from the simplest Type-1 (observation-driven Markov decision process) to Type-5 (DOSASE MDP), are introduced. Further, we present the architecture design of the Dav robot. The framework of the Dav architecture is hand-designed, but the actual controller is developed, i.e., generated autonomously by the developmental program through real-time interactions with the real physical environment. We present the Dav architecture and the major components that implement the architecture. (3) Based on the above developed robotic platform and the proposed architecture, we have designed and implemented a range-based navigation system. |
