Distributed optimization system
AssigneeSandia Corporation
InventorsJohn E. Hurtado, Clark R. Dohrmann, Rush D. Robinett III
Life of this patent
Estimating the term…
Abstract
A search system and method for controlling multiple agents to optimize an objective using distributed sensing and cooperative control. The search agent can be one or more physical agents, such as a robot, and can be software agents for searching cyberspace. The objective can be: chemical sources, temperature sources, radiation sources, light sources, evaders, trespassers, explosive sources, time dependent sources, time independent sources, function surfaces, maximization points, minimization points, and optimal control of a system such as a communication system, an economy, a crane, and a multi-processor computer.
Claims24
1) A distributed search system for controlling a plurality of agents to search an area for an objective, wherein the distributed search system comprises an agent search system associated with each agent, and wherein each agent search system comprises:
- a) a sensor;
- b) a communicator, responsive to the sensor and communicating a sensor reading;
- and
- c) a cooperative controller, responsive to sensor readings from some of the plurality of agents and generating an agent control strategy determined from the sensor reading and a cooperative approximation to the search area.
14) A method for controlling a plurality of agents to search an area for an objective, wherein each agent comprises a sensor, a communicator, and a cooperative controller, the method comprising on each agent:
- a) determining a reading from the sensor;
- b) sharing the reading with other agents in the plurality of agents;
- c) concurrently approximating the search area, according to a plurality of readings from the plurality of agents;
- d) solving for a cooperative approximation, corresponding to the concurrently approximated search area;
- e) updating the agent control strategy, according to the cooperative approximation and the reading;
- and
- f) repeating steps a) through e) until an optimization condition for the objective is met.
23) A method for controlling a plurality of processors in a computer to cooperatively optimize an objective for a surface function, the method comprising on each processor:
- a) evaluating the surface function at a location;
- b) sharing the evaluation and the location with other processors in the plurality of processors;
- c) concurrently approximating the surface function, according to evaluations and locations from the plurality of processors;
- d) solving for a cooperative approximation, corresponding to the concurrently approximated surface function;
- e) updating a processor control strategy, according to the cooperative approximation and the position;
- and
- f) repeating steps a) through e) until an optimization condition for the objective is met.
Dependencies are inferred from the claim text.