Design of a Two-Level Fuzzy Controller for a Reactive Miniature Mobile Robot

This paper describes the design and implementation of the fuzzy control system for a small, reactive mobile robot which operates in an unknown , unpredictable, and dynamic environment. A modular, two-level fuzzy controller is used for navigation, obstacle avoidance, and target tracking. The fuzzy controller provides the mechanism for fusing noisy sensor data from multiple sensors which may present connicting information. Competing behaviors of target tracking and obstacle avoidance are combined in a way that serves both functions, thereby resulting in an emergent intelligent behavior of the robot controller. The ee-ciency of the two-level fuzzy controller is demonstrated by implementing the system on a small mobile robot with two onboard processors and limited memory (2K per processor). Real time navigation of mobile robots in dynamic, un-structured environments requires eecient processing of large amounts of sensory data. Sensor data may be noisy or incomplete. Computing resources may be limited. In spite of this, the robot must perform decision-making and reasoning about its surroundings. To handle the challenges of real time sensory input and processing in an unknown and constantly changing environment, reactive architectures have been used. One reactive approach is the subsumption behavior architecture developed by Brooks 2, 3]. Independent, decision-making processes, called behaviors, execute concurrently. Each behavior receives sensor data relevant to its decision-making needs and produces a control output based on its desired action. Independent behaviors may present connicting control actions that must be resolved. Another reactive approach uses schemas to connect sensor input and control output 1]. Likewise, outputs from multiple schemas may have to be combined or reconciled to produce a nal control action. Resolving connicting control outputs can present problems , as discussed in 5]. Recently, researchers have applied fuzzy logic to the reactive behaviorist approach 8, 7, 6]. The advantage to using fuzzy logic is that potentially connicting functions can be fused in a natural way, so that a reasonable decision can be made to serve both functions. In the fuzzy logic behaviorist systems, special emphasis is placed on ensuring independence between behavior modules. Independent fuzzy inference units are used for each behavior. This approach has the advantage of reducing the complexity; however, it ignores any natural coupling that may exist between control actions of behaviors. Our approach addresses the coupling problem, while at the same time minimizes the complexity. Our objective was to create a small mobile robot, completely autonomous and purely reactive, using …