• Home »
• Adaptive Robotics

The interface allows users to designate an area to be searched.

In order to create behaviors that can be used in dynamic, unknown environments, the INL has emphasized a tight relationship between sensing and action rather than attempting to model the world as the robot explores it. Each of the sensors on the robot feeds into an array of regions centered on the robot. The robot responds to this sensor information based on fuzzy logic rules that compete for control of the robot’s rotational and translational velocity. These fuzzy logic rules not only allow the robot to respond to each individual region, but also can be triggered by combinations and patterns found within the array of regions.

The INL has designed an autonomous control architecture that includes a variety of reactive and deliberative behaviors. Reactive behaviors are mappings between sensing and action, such as: obstacle avoidance; maneuver; get-unstuck; and follow. The INL uses a subsumption scheme such that each reactive behavior can override the output of other behaviors. The concept of subsumption architecture was originally developed by Brooks in 1986; it provides a method for structuring reactive systems from the bottom up using layered sets of rules. Within the INL’s architecture, each reactive behavior runs independently. Behaviors such as obstacle avoidance may run continuously, supporting a spectrum of reactive and deliberative capabilities that operate in parallel. Although this subsumption architecture provides a strong foundation of basic capabilities, INL has also incorporated deliberative behaviors, which exploit a world model and function at a level above the reactive behaviors by combining the abilities afforded by the reactive behaviors. Once the reactive behaviors are “satisfied,” the deliberative behaviors may take control, allowing the robot to accomplish high-level behaviors such as “area search,” “patrol perimeter,” and “follow route.”