P.E. Agre. The Dynamic Structure of Everyday Life. PhD dissertation, MIT Artificial Intelligence Laboratory, Cambridge, MA, 1988.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....the counterfactual reasoning of planning as theory extension . Their implemented system TEST [106] made explicit some of the assumptions made by planners, as well as generalising the control strategy of TWEAK to encompass interval based representations of time. Chapman and Agre s PENGI programme [2, 1] is an attempt to generate the kinds of purposeful action normally associated with plans, by using rules to take advantage of whatever the world affords locally rather than a representation of the world changing over time. Chapman s own thesis [21] described an extension of such a system to ....

Philip E. Agre. The Dynamic Structure of Everyday Life. PhD thesis, MIT Artificial Intelligence Laboratory, 1988. Available as Technical Report MIT-TR-1085.

....of the domain features into the categories of fast reaction and limited knowledge, this discussion will follow the same division. 2.2. 2 Reactive planners A prime candidate for the any domain that requires rapid reactions is the research that falls under the heading of reactive planning [1, 5, 6, 34]. These systems were designed specifically to react rapidly to changes in the environment. These systems calculate the next action on the basis of the system s current inputs. To do this the system designer determines the correct actions for each possible set of inputs when building the system. ....

....the system is missing may be crucial for the system to retrieve the correct plan, missing even a single fact can cause the system to retrieve an inappropriate plan that will have to be extensively modified sacrificing the speed of the planning process as well. 6. 4 Reactive planning Agre [1, 2], Brooks [5] Chapman [7] Kaelbling [18] and others have argued that rather than deliberative processes, behavior in autonomous agents can best be created by simple mappings from sensor inputs to action outputs. These sensor action pairs can be layered and interleaved in order to calculate ....

Philip Agre. The dynamic structure of everyday life. Technical Report 1085, MIT Artificial Intelligence Laboratory, 1988.

....language to talk about conditional effects of actions but does not provide a practical solution for how to plan using them. It also rejects the fundamental realization of action in context, namely: an action has no effects outside of the environment in which it is performed. 1 Other researchers [1, 2, 12] have solved the problems of the context sensitivity of action effects by considering all possible world states and deciding before hand the correct action. While these systems do work, once built they lack flexibility since the correct action cannot be changed at runtime. They have traded ....

P. Agre. The dynamic structure of everyday life. Technical Report 1085, MIT Artificial Intelligence Laboratory, 1988.

....a number of researchers have proposed theories of agent environment 36 interaction that have been called reactive or reaction systems. These systems stress the need for an agent to be able to respond to the world as rapidly as possible. Pengi, Sonja, and Rex Agre, Chapman, Kaelbling, and others [1, 2, 5, 19] have argued that the most important ability an agent can have is the ability to respond rapidly to its environment. To this end, they are willing sacrifice actual planning. Essentially they encode the desired responses of the agent as a function from sensor inputs to effector outputs. The ....

Philip Agre. The dynamic structure of everyday life. Technical Report 1085, MIT Artificial Intelligence Laboratory, 1988.

....the state of the world and its memory to determine if any of the goals within its plan have been satisfied. When a goal has been satisfied serendipitously, it can be pruned out of the plan structure, and the system can move on to consider its next goal. Second, ItPlanS leans on the world [1] when predicting the results of its actions. Rather than maintaining a complete model of the world and the state that results from executing the action, ItPlanS uses a simpler method based on associating conditional add and delete lists with each action. ItPlanS assumes that a given proposition is ....

Philip Agre. The dynamic structure of everyday life. Technical Report 1085, MIT Artificial Intelligence Laboratory, 1988.

....its environment. This parallels Brooks approach whose lowest layer of control insures that his robots do not come into contact with other objects: they achieve zero level competence [28, 29, 30, 32] For a good discussion on avoidance concerns and techniques, see Reynolds [132] Both Agre [2] and Wehner [155] observe that animals do not have guaranteed general purpose collision avoidance strategies. Instead they use simple basic avoidance techniques that are effective most of the time in concert with a small set of special purpose avoidance techniques for specific cases. When ....

Philip Agre. The Dynamic Structure of Everyday Life. PhD thesis, MIT Artificial Intelligence Laboratory, 1988. Technical Report 1085.

....the state of the world and its memory to determine if any of the goals within its plan have been satisfied. When a goal has been satisfied serendipitously, it can be pruned out of the plan structure, and the system can move on to consider its next goal. Second, ITPLANS leans on the world [1] when predicting the results of its actions. Rather than maintaining a complete model of the world and the state that results from executing the action, ITPLANS uses a simpler method based on associating conditional add and delete lists with each action. ITPLANS assumes that a given proposition ....

P. Agre. The dynamic structure of everyday life. Technical Report 1085, MIT Artificial Intelligence Laboratory, 1988.

....designs that were not thought of originally. While we hope that the simulator will be generally useful, our design was also influenced by recent work in reactive planning, particularly Firby s RAP system [4] and McDermott s RPL language [7] as well as Agre s work on deictic representation [1]. 1 Abstract Realistic Simulation of Mobile robots for Analyzing Goal achievement, N avigation and Adaptation 2 CHAPTER 1. INTRODUCTION 1.1 Simulated Domains When designing an environment for experimentation in planning, there are a number of issues of that must be addressed. There are three ....

....through the function robot camera: robot camera robot index 3.4. 1 Designators The purpose of designators in the system is to provide a sort of indexical naming of objects in the system based on the notion of effective designators developed in [7] see also deictic representation discussed in [1]) A designator provides a handle on an object or place (its denotation) that the system can use to control the robot, relative to the designator s denotation. In the simulator there are two distinct types of designators, local and non local. Local designators denote objects next to the robot, ....

[Article contains additional citation context not shown here]

Philip E. Agre. The Dynamic Structure of Everyday Life. PhD thesis, MIT Artificial Intelligence Laboratory, 1988.

....only the events considered at planning time will happen at plan execution time. Reaction planning tackles the problem by (sometimes automatically) encoding goal directed behavior not only for a single planning problem but for whole parts of the problem domain (e.g. Georgeff and Lansky, 1987; Agre, 1988; Schoppers, 1989b ] Reactive planning follows the idea of providing rough plans that are used, reacting to unforeseen or unforeseeable events in the world at execution time (e.g. Dean and Boddy, 1988 ] If you agree with the reasons for working on the individual topics, then it seems ....

P.E. Agre. The Dynamic Structure of Everyday Life. PhD thesis, MIT Artificial Intelligence Laboratory, 1988.

....without representation Rodney A. Brooks MIT Artificial Intelligence Laboratory, 545 Technology Square, Rm. 836, Cambridge, MA 02139, USA Received September 1987 Brooks, R.A. Intelligence without representation, Artificial Intelligence 47 (1991) 139 159. This report describes research done at the Artificial Intelligence Laboratory of the Massachusetts Institute of Technology. Support for the research is provided in ....

....without representation Rodney A. Brooks MIT Artificial Intelligence Laboratory, 545 Technology Square, Rm. 836, Cambridge, MA 02139, USA Received September 1987 Brooks, R.A. Intelligence without representation, Artificial Intelligence 47 (1991) 139 159. This report describes research done at the Artificial Intelligence Laboratory of the Massachusetts Institute of Technology. Support for the research is provided in part by an IBM Faculty 9 ....

[Article contains additional citation context not shown here]

P.E. Agre and D. Chapman, Unpublished memo, MIT Artificial Intelligence Laboratory, Cambridge, MA (1986).

....restrictive assumptions are violated. To this list, we would like to add one more: planning s view of activity is essentially one shot. That is, classical planning techniques produce a plan to satisfy a given goal set, given a particular situation. There is no place for time extended interaction ([1]) with an environment: presumably either a planner plans for the whole of time in advance, or is invoked at appropriate times to deal with particular goal sets. The former alternative multiplies the intractability of planning, while the latter requires, at the minumum, a theory of when to invoke ....

....that is relevant is defined in terms of single tasks, rather than in interaction with an environment over time. 1.2. 2 Situated Action models Discontents with planning models have led to an enormous variety of research in recent years in this section we will focus on situated action models [1, 4], and leave connections with closely related work to the reader. This line of work stresses the interaction of agents and their environments, and their mutual dependence. It argues that design of an all purpose agent is not feasible, and that it is incumbent on the designer of an agent to ....

[Article contains additional citation context not shown here]

P. E. Agre. The Dynamic Structure of Everyday Life. PhD thesis, MIT Artificial Intelligence Laboratory, 1988. Technical Report 1085.

....the same as that of PRS the lack of a low level control layer. A completely different approach to planning was developed by Agre and Chapman of MIT. Their system was based on routines, which are methods of representing objects in the world based on the interaction of the robot and those objects [1, 2, 3]. The idea behind routines is that a human s usual activities are not symbolic reasoning or planning, but simply carrying out pre learned routines. In this view, plans allow some improvisation, much like 3.5. LIMITATIONS OF PREVIOUS WORK 24 a cook uses a recipe. Indeed, the cook could perform ....

P. E. Agre, "Routines," Technical Report MIT AIM828, MIT Artificial Intelligence Laboratory, Cambridge, MA, May 1985.

....of setting, the agent often does not care to uniquely identify objects. It is sufficient to know the current relationship of the relevant objects to the agent, and what roles the objects play in the agent s activities. Agre and Chapman in [AC87] proposed indexical functional representations (which [Agr88] refers to as deictic representations) to be the more natural way agents refer to objects in common everyday environments. They called entities and relationships of interest entities and aspects, respectively. With respect to its current activities, the agent needs only to focus on representing ....

....control. Maes has explored learning and has applied her architecture to robotic systems. In the subsumption architecture, sensations and actions are abstracted by giving them names like straightening behavior in order to make things easier to understand for human observers. Much in the spirit of [Agr88], we believe that behavior modules should more naturally emerge from the interaction of the agent with its environment. In contrast to hand coding behaviors and in order to facilitate embodiment, in GLAIR we are experimenting with (unnamed) emergent behavior modules that are learned by a robot ....

Philip Agre. The dynamic structure of everyday life. Technical Report 1085, MIT Artificial Intelligence Laboratory, MIT, 1988.

.... such as color, size, and shape, than on interactional properties and relationships, such as graspable and fits in my mouth, which characterize how an agent interacts with its environment [ Johnson, 1987; Lakoff and Johnson, 1980; Lakoff, 1984 ] At the same time, AI researchers such as Agre [ Agre, 1988 ] Chapman [ Chapman, 1991 ] and Ballard [ Ballard, 1989 ] have argued for deictic or agent centered representations. Lakoff and Johnson make a convincing case that the primitive interactional knowledge acquired by infants becomes more elaborate through abstraction and metaphorical extension ....

Philip E. Agre. The dynamic structure of everyday life. Technical Report 1085, MIT Artificial Intelligence Laboratory, Cambridge MA, 1988.

....of setting, the agent often does not care to uniquely identify objects. It is sufficient to know the current relationship of the relevant objects to the agent, and what roles the objects play in the agent s activities. Agre and Chapman in [2] proposed indexical functional representations (which [1] refers to as deictic representations) to be the more natural way agents refer to objects in common everyday environments. They called entities and relationships of interest entities and aspects, respectively. With respect to its current activities, the agent needs only to focus on representing ....

....control. Maes has explored learning and has applied her architecture to robotic systems. In the subsumption architecture, sensations and actions are abstracted by giving them names like straightening behavior in order to make things easier to understand for human observers. Much in the spirit of [1], we believe that behavior modules should more naturally emerge from the interaction of the agent with its environment. In contrast to hand coding behaviors and in order to facilitate embodiment, in GLAIR we are experimenting with (unnamed) emergent behavior modules that are learned by a robot ....

P. Agre. The dynamic structure of everyday life. Technical Report 1085, MIT Artificial Intelligence Laboratory, MIT, 1988.

....ice cream implicitly means, go find a scoop and use it to scoop the icecream. This suggests that if we are going to design robotic agents to assist us, they must be able to locate objects and manipulate them. For many planning systems, the issue of searching for objects never arises. For example, [1, 5, 12, 17] work under the simplifying assumption that the agent knows all of the objects in the world and their locations, and every object in a plan uniquely refers to an object in the world. Thus when the robot is given the command Get a scoop, the command actually refers to a unique scoop; the assistant ....

Philip Agre. The dynamic structure of everyday life. Technical Report 1085, MIT Artificial Intelligence Laboratory, 1988.

No context found.

P.E. Agre. The Dynamic Structure of Everyday Life. PhD dissertation, MIT Artificial Intelligence Laboratory, Cambridge, MA, 1988.

No context found.

Philip Agre. TheDynamic Structure of Everyday Life. PhD thesis, MIT Artificial Intelligence Laboratory, 1988. Technical Report 1085.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC