BBN Technologies

In a wide range of computational areas, arriving at the optimal solution (or even an acceptable solution) can be a time-consuming task. Whether you are working in planning, design, scheduling, dynamic adaptation, or other problem-solving domains, arriving at the best outcome using software that employs traditional means can require long, inefficient efforts at applying a large set of potential solutions.

The answer is better software-engineering algorithms - ones that provide a faster, more efficient approach to both software development and application execution - algorithms that cost less to develop, that arrive at better outcomes, and that can provide a more robust result in the bargain.

BBN has developed just such an approach to computational problem solving; an approach that that optimizes outcomes by employing "genetic algorithms," (and their specialized cousins, "genetic programming" techniques). These optimize outcomes by applying the same processes of adaptation that we find in natural evolution. Why use this approach? Because it can be exceedingly fast while analyzing a large number of potential solutions, and it can uncover solutions that might not even occur to a human analyst.

BBN has created optimizing applications using this approach in all of the following areas:

• Scheduling
  BBN has created automated scheduling systems for allocating commercial field service personnel, scheduling military transportation aircrews, and scheduling computational resources in a distributed-processing environment. Our automated scheduler, called Vishnu, features scheduling logic that can be configured for most scheduling problems without modifying the software.
• Robot Mission and Path Planning
  In the area of robotic intelligence, BBN's genetic algorithms can identify in real time a nearly optimal set of actions to accomplish a given mission.
• Ad Hoc Network Parameter Optimization
  For a mobile, ad hoc, networking problem BBN used a genetic algorithm to tune the parameters of a protocol to the characteristics of both the nodes and the environment.
• Communication Network Design
  To allocate satellite and wireless bandwidth in a network design, BBN arrived at an optimal solution by applying our expertise in genetic algorithms.
• Behavior Learning for UAVs
  For optimizing mission strategies for unmanned aerial vehicles (UAVs), BBN has used genetic programming (a type of genetic algorithm).
• Sonar Processing Parameter Optimization
  To tune the parameters of sonar-processing algorithms to the characteristics of the underwater targets and environment, BBN used a specialized, genetic-algorithm approach.
• Automated Learning for Control of Networked Traffic Signals
  In the area of adaptation and learning, BBN has shown how genetic programming can teach a system to learn control strategies for traffic signals in a network of streets.