A strategy for autonomous navigation of a robotic vehicle across hazardous terrain involves the use of a measure of traversability of terrain within a fuzzy-logic conceptual framework. This navigation strategy requires no a priori information about the environment. Fuzzy logic was selected as a basic element of this strategy because it provides a formal methodology for representing and implementing a human driver's heuristic knowledge and operational experience.

Image Data Acquired by Onboard Stereoscopic Cameras is processed into terrain-quality information and summarized in a fuzzy traversability index.

Within a fuzzy-logic framework, the attributes of human reasoning and decision-making can be formulated by simple IF (antecedent), THEN (consequent) rules coupled with easily understandable and natural linguistic representations. The linguistic values in the rule antecedents convey the imprecision associated with measurements taken by sensors onboard a mobile robot, while the linguistic values in the rule consequents represent the vagueness inherent in the reasoning processes to generate the control actions. The operational strategies of the human expert driver can be transferred, via fuzzy logic, to a robot-navigation strategy in the form of a set of simple conditional statements composed of linguistic variables. These linguistic variables are defined by fuzzy sets in accordance with user-defined membership functions. The main advantages of a fuzzy navigation strategy lie in the ability to extract heuristic rules from human experience and to obviate the need for an analytical model of the robot navigation process.

The basic building block of the present navigation strategy is a behavior, defined here as a representation of a specific sequence of actions aimed at attaining a given desired objective. Each behavior comprises a set of fuzzy-logic rules of the form

IF C, THEN A,

where the condition C is composed of fuzzy input variables and fuzzy connectives (AND, OR, NOT), and the action A is a fuzzy output variable. Such an IF, THEN rule represents a typical rule in a set of natural linguistic rules that express the actions taken by an expert human driver based on the prevalent conditions. The output of each behavior describable by such a rule set is a recommendation over all possible control actions from the perspective of attaining the objective.

Multiple behaviors, each aimed at one specific goal, can be active simultaneously in the navigation strategy. Blending of multiple behaviors is implemented by combining the outputs (recommendations) of all the behaviors using gain rules of the form

IF S, THEN K,

where S is a logical statement that describes a physical situation, and K represents a fuzzy expression of the gains with which the recommendation of the individual behaviors are weighted in the prevalent situation. The result of the weighted combination of recommendations is then issued as a command to the wheel actuators of the mobile robot.

The present robot-navigation strategy involves three such behaviors, denoted seek-goal, traverse-terrain, and avoid-obstacle:;

  • The navigation rules for the seek-goal behavior utilize the global information about the goal position to generate the steering and speed commands that drive the robot to the designated destination.
  • The navigation rules for the traverse-terrain behavior utilize regional information about the quality of the terrain to produce steering and speed commands that guide the robot toward the safest and the most traversable terrain. The regional terrain-quality information is generated from readings of onboard sensors by use of a set of fuzzy-logic rules. This behavior constitutes a major novel aspect of the present strategy (see figure).
  • The navigation rules for the avoid-obstacle behavior employ local information about obstacles en route to develop steering and speed commands to maneuver the robot around the obstacles.

The recommendations of these three behaviors are blended through gains or weighting factors to generate the final steering and speed commands to be executed by the wheel actuators of the robot. The gains are also generated by fuzzy-logic rules that take into account the current status of the robot.

This work was done by Homayoun Seraji, Ayanna Howard, and Bruce Bon of Caltech for NASA's Jet Propulsion Laboratory. For further information, access the Technical Support Package (TSP) free on-line at www.nasatech.com/tsp under the Information Sciences category.

This software is available for commercial licensing. Please contact Don Hart of the California Institute of Technology at (818) 393-3425. Refer to NPO-21199.