Using a Blackboard to Integrate Multiple Activities and Achieve Strategic Reasoning for Mobile-Robot Navigation

• Authors: Ramiro Liscano;Reda E. Fayek;Allan Manz;Elizabeth R. Stuck;Jean-Yves Tigli
• Venue: IEEE Expert: Intelligent Systems and Their Applications
• Year: 1995

The activity-based blackboard system consists of twohierarchical layers for strategic and reactive reasoning: Ablackboard database to keep track of the state of the world and aset of activities to perform real-time navigation.Developing a mobile robot gives us the opportunity toinvestigate issues in the design of intelligent systems because therobot's mobility forces us to deal with many unpredictableenvironmental situations. One dictionary definition of intelligenceis the ability to deal with new or trying situations. Thus, amobile robot that reliably navigates in unknown environments givesthe appearance of intelligent behavior. The main idea of anautonomous vehicle is quite simple: Given a task to perform, itmust have the ability to perceive the environment and actappropriately. This ability requires a feedback control system tolink the vehicle's sensing and control. Unfortunately, autonomousrobots have characteristics not yet satisfactorily addressed by theclassical control community:Solving the problems encountered by the mobile robot generallyrequires the integration of several methodologies.The robot's decision space is discrete and composed of distinctelements as opposed to continuous functions.The system must react to the environment in an appropriate timeperiod.Due to the limitations of the sensors and sensory processing,most of the knowledge the robot acquires is either incomplete oruncertain.To address these problems, we have designed a mobile-robotsystem architecture that uses a blackboard to coordinate andintegrate several real-time activities. An activity is anorganizational unit, or module, designed to perform a specificfunction, such as traversing a hallway, going down steps, crossingover an open channel on the floor, or tracking a landmark. Anactivity resembles a behavior in that it controls the robot toperform a specific task. It differs from a behavior in that it isdesigned to perform the specific task in a narrow applicationdomain, whereas a behavior generally resembles a biologicalresponse--that is, an organism's response to a stimulus.Payton defined the term activity as an instance of an activationset, where an activation set is composed of a number of behaviors.Payton's activities are a way to specify a combination of behaviorsto achieve a more complex behavioral pattern. In contrast, we makeno attempt to define our activities as a combination of basicbehaviors. In our system, several activities are necessary for therobot to perform simple tasks such as moving around a factory bay.Some of these basic navigation activities are traversing openspace, crossing over floor anomalies (cables or channels), andavoiding collisions.The system architecture must define a mechanism to coordinatethe mobile robot's activities since they cannot all drive the robotsimultaneously. Most mobile-robot control systems are hybridsystems combining approaches from hierarchical, behavioral, andblackboard-based systems. Behavior-based systems have recentlybecome more prevalent for controlling mobile robots.Hierarchical architectures, such as NASREM (NASA/NIST StandardFunctional Architecture for Telerobot Control System Architecture)and IMAS (Intelligent Mobile Autonomous System), offer a niceparadigm for breaking down a global task into subtasks, but thehierarchical structure quickly degrades into a hierarchical commandstructure combined with distributed sensing similar to thatimplemented on Mobot III. Most of these systems use sense-think-actcycles that are difficult to implement in real time when the robotmust deal with diverse sensing conditions. Lumia has demonstratedthe use of NASREM-style architecture for real-time tracking andcatching a ball falling through a maze of pegs, but this system hasa well-defined, narrow scope of operation and does not require thesensing diversity that a mobile robot needs.In an attempt to add reflexive ability to a hierarchical system,Payton proposed a vertical decomposition along with the moreclassical horizontal, hierarchical decomposition, resulting in ahierarchical structure composed of reflexive behaviors at thelowest level. The effect is a hierarchical system capable ofmanaging diverse sensing conditions--and therefore a more robustsystem. Arkin also emphasized the importance of a nonhierarchicalbroadcast of information. Arkin chose to design a system usingagents that manipulate a unified representation of the world basedon potential fields. Our system does not use a unifiedrepresentation to coordinate activities but relies on proceduralknowledge and sensory data posted on the blackboard to make adecision.Similar to the Codger system architecture used to drive CMU'sNavlab, our system does not use its blackboard as a problem-solvingmechanism but primarily as the supervisor and coordinator ofseveral real-time activities. These activities continually posttheir current state and the current state of the environment to theblackboard. The perception and sensing components of an activityare designed to run concurrently. Not all communications between anactivity's modules go through the blackboard, thus reducingcommunications bandwidth and making reactive behavior possible.Like a traditional blackboard system, our system uses a centraldatabase to store information accessible by a number of modules,but it differs functionally because the modules do not workcooperatively to solve a common problem.In our system architecture, the blackboard's rule set andknowledge determine which activity controls the mobile robot'sactuators. A production system facilitates experimentaldetermination of adequate conditions for selecting the activitycontrolling the vehicle, and the blackboard database serves as arepository of state data and sensory data from the activities.Strategic reasoning is the ability to process sensory inputs,stored information, and long-term goals so that the robot can makedecisions with a global view of the environment. The decision aboutwhich activity will control the vehicle is based on sensory andstate information from the activities and therefore is a form ofstrategic reasoning.