DCOPs Meet the Real World: Exploring Unknown Reward Matrices With Applications to Mobile Sensor Networks

Free registration required

Executive Summary

Buoyed by the recent successes in the area of Distributed Constraint Optimization Problems (DCOPs), this paper addresses challenges faced when applying DCOPs to real-world domains. This expedition reveals that three fundamental challenges must be addressed for a large class of real-world domains, requiring design of novel DCOP algorithms. First, in many domains, agents do not know the initial payoff matrix and must explore the environment to determine rewards associated with different variable settings. Second, the agents have a goal to maximize the total accumulated reward rather than the instantaneous reward at the end of the run. Third, limited task-time horizons disallow agents the luxury of full exploration of their environment and payoff matrices.

  • Format: PDF
  • Size: 451 KB