Reactive Behavior

Abstract

We are now ready to write our first algorithms for robots. These algorithms demonstrate reactive behavior: an event (such as the detection of a nearby object by the robot) causes the robot to react by performing an action that changes its behavior (such as stopping the motors). Purely reactive behavior occurs when the action is related only to the occurrence of an event and does not depend on data stored in memory (state). The reactive behaviors are those of Braitenberg vehicles which are appropriate for introducing robotics because complex behavior arises from simple algorithms. Sections 3.1-3.3 describe Braitenberg vehicles that demonstrate reactive behavior; in Chap. 4 we present Braitenberg vehicles that have non-reactive behavior with states. Section 3.4 presents several algorithms for the classic reactive behavior of line following. Line following is an interesting task because the algorithms are sensitive to the characteristics of the sensors and the lines. Calibration is necessary to determine the optimum thresholds for fast robust motion of the robot. Section 3.5 gives a brief overview of Braitenberg's original formulation of the vehicles in a biological approach with sensors connected directly to the motors, not through a computer. Later in the book (Sect. 13.3) we discuss the implementation of Braitenberg vehicles using neural networks. 3.1 Braitenberg Vehicles Valentino Braitenberg was a neuroscientist who described the design of virtual vehicles that exhibited surprisingly complex behavior. Researchers at the MIT Media Lab developed hardware implementations of the vehicles from programmable bricks that