"> Project Goals

Project Goals

Free Market Architecture
  1. Formalize the core components of a free market architecture.
  2. Implement architecture and evaluate performance in simulation.
  3. Extend architectural capability to accomodate exploration of partially known worlds.
  4. Extend architectural capability to accomodate robot death.
  5. Evaluate architectural performance on real test bed with four robots.
  6. Extend architectural capability to accomodate multiple roles.
Map Reconstruction
  1. Robust matching of features between images taken at arbitrary positions by different robots and recovery of robot poses and scene geometry.
  2. Recovery of dense depth maps from images from robots at arbitrary positions.
  3. Generation of texture-mapped realistic environment model from the reconstruction for presentation to users.
  4. Integration of multiple maps into a combined representation of the environment.
  5. Use of the geometric representation for reasoning about occlusions and view planning.
Communal Learning
  1. Develop communal learning techniques in which the colony shares experiences (data gathering) and the computational burden of learning.
  2. Prove that concepts can be learned by the colony that are beyond the capabilities of the individual.
  3. Show the benefit of having learned information preserved across the colony. Also show that the colony can reuse and build upon previously learned skills and behaviors. In situations, where new generations of robots replenish those consumed, this will lead to a progressive society where new generations can advance beyone the previous generation.
  4. Develop a "buddy" system for insuring fatal state information survives. Communal learning can then learn fatal situations and actions.
  5. In many situations, robot sacrifice will be required. Techniques will be developed that will insure that the most information is gained for the colony from any sacrifice.
  6. Fit communal learning into the free market architecture. By make experiences, computation, and new or refined behaviors/information a commodity, communal learning will be part of distributed market architecture.
  7. Show that communal learning can lead to more accurate and meaningful negotiations. This will be done by learning transaction confidence and expected costs.
  8. Apply communal learning to: mobility models, behavior parameters, transaction models, maps and enemy tactic models for individuals and teams.
Robot Test Bed
  1. 30 cm/s reliable obstacle navigation.
  2. Integration of Firewire hardware on each platform.
  3. Integration of audio hardware on each platform.
  4. Data Abstraction Module for image acquisition (both streaming and static forms) and audio clips.
  5. Multi-link routing module that enables point-to-point communication through multiple intermediaries to compensate for limited range in our wireless hardware.
  6. Augment sonar-based obstacle avoidance with image-based approach.
  7. Characterize odometric data (linear and angular drive vs time).
  8. Develop network resource manager that empowers robots to obtain information and physical services from others.
  9. Design and integrate mounting brackets for audio and imaging hardware.