A subsumption planner using a parallel distributed computational paradigm based on the subsumption architecture for control of real-world capable robots is described. Virtual sensor state space is used as a planning tool to visualize the robot's anticipated effect on its environment. Decision sequences are generated based on the environmental situation expected at the time the robot must commit to a decision. Between decision points, the robot performs in a preprogrammed manner. A rudimentary, domain-specific partial world model contains enough information to extrapolate the end results of the rote behavior between decision points. A collective network of predictors operates in parallel with the reactive network forming a recurrrent network which generates plans as a hierarchy. Details of a plan segment are generated only when its execution is imminent. The use of the subsumption planner is demonstrated by a simple maze navigation problem.


    Access

    Access via TIB

    Check availability in my library


    Export, share and cite



    Title :

    Planning in subsumption architectures


    Contributors:


    Publication date :

    1994-03-01


    Type of media :

    Conference paper


    Type of material :

    No indication


    Language :

    English


    Keywords :