Creative Robots to Defeat Deeply Buried Targets

Period of Performance: 06/11/2003 - 03/11/2004


Phase 1 SBIR

Recipient Firm

Imagination Engines, Inc.
12906 Autumn View Dr
St. Louis, MO 63146
Principal Investigator


Recent developments in the area of artificial neural networks have led to a totally new brand of machine intelligence that is capable of autonomously improvising and implementing cunning concepts and strategies. This revolutionary neural network technology, called the "Creativity Machine Paradigm," has led to a wide variety of clever, self-learning neural architectures that will inevitably enable a whole new generation of truly autonomous robots. Designed as OOP class templates, these self-learning Creativity machines may be instantiated into thousands of independent or cooperative agents. The resulting intelligent swarms may exhaustively learn from their digital or real-world environments, individually or collectively manipulating their surroundings. Each of these objects may evolve suitable sensors and effectors to best achieve their objectives. Self-perfected, this new form of artificial life may be harvested from the simulation environment and then implemented within hardware through ASIC or reconfigurable computing schemes. Herein we propose to examine the feasibility of designing and testing these advanced connectionist paradigms within a specialized integrated development environment to aid in CONOPS development. The highly matured parallel algorithms derived from these simulations would be converted to hardware based control systems for use in miniaturized attack robots specialized to penetrate deeply buried underground facilities. Technology developed on this program will be applicable to commercial and military robots and consumer toys. Spin-off from this effort would result in major advances in self-configuring electronic devices, micro-surgical robots, autonomous control systems, as well as the real-time generation of physically plausible virtual environments, whether for entertainment or tactical simulations. Furthermore, each subtask of this proposed Phase I effort will generate products and services having solid commercial viability. Among these numerous by-products will be (1) neural models of sensors, actuators, and electronic interfaces that allow virtual performance testing to prospective buyers, (2) a general development environment for advanced neural network cascade construction, (3) educational toys for the investigation of artificial life, and (4) a development environment for battlebots that may be applied toward military or civilian ends.