A. Ram and J. C. Santamaria. Continuous case-based reasoning. Artificial Intelligence, 90(1-2):25--77, 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....problem domain. If the solution adapted for the current problem is closer to some other case than the template, then it might be so that the template s regularities are not present in the current problem. This approach bears some philosophical resemblance to the methods used by Ram and Santamaria [15] for case based robotic control. Our implementation of these ideas, however, is quite different from theirs. 2.4. Computational complexity The time complexityof the recipe planning is mainly affected by the case retrieval and adaptation generation. The retrieval cost is incurred in the pair wise ....

A. Ram and J.C. Santamaria, Continuous case-based reasoning, Artificial Intelligence 90 (1996), 25--77.

.... 1997] Other PRIAR [ Kambhampati and Hendler, 1992 ] Other PRODIGY ANALOGY [ Veloso, 1994; Veloso et al. 1995; Veloso et al. 1996 ] Logistics ROBBIE [ Fox and Leake, 1995a; Fox and Leake, 1995b ] Navigation Planning ROUTER [ Goel et al. 1994; Goel et al. 1995 ] Navigation Planning SINS [ Ram and Santamar ia, 1997 ] Navigation Planning SPA [ Hanks and Weld, 1995] Other TOLTEC [ Tsatsoulis and Kashyap, 1993 ] Process Planning Table 1: Case based planning systems and their application domains. 9 Initial Situation New Goal and General Domain Adapted Plan Plan Knowledge Learned Plan Tested Repaired ....

A. Ram and J.C. Santamar'ia. Continuous Case-Based Reasoning. Artificial Intelligence, 90(1--2):25--77, 1997.

.... for a machine vision system that recognizes hand gestures (Darrell, Essa, Pentland 1996) for a simulated agent that maneuvers along a highway (McCallum 1996) and for a human computer interface that automates repetitive tasks (Das, Caglayan, Gonsalves 1998) Like Pierce and Kuipers (1997) Ram and Santamaria (1997) and others, e.g. Iba 1991; Thrun 1999) we believe that sensorimotor agents can discover categories for themselves. Thus, the focus of this paper is an unsupervised method by which a mobile robot deduces meaningful categories from uninterpreted sensor readings. Previously, we demonstrated a ....

....sensor readings. From Sensors to Categories For a mobile robot operating in an environment of even modest complexity, sensory categories supply a needed level of abstraction away from raw sensor readings (Mahadevan, Theocharous, Khaleeli 1998; Michaud Mataric 1998; Pierce Kuipers 1997; Ram Santamaria 1997). Since our objective is that agents discover such categories for themselves without supervision we make use of clustering techniques that o er a general, unsupervised framework for categorizing data. However, the following subproblems exist, and this section outlines our solution to each one: ....

[Article contains additional citation context not shown here]

Ram, A., and Santamaria, J. C. 1997. Continuous case-based reasoning. Articial Intelligence 90:25-77.

....acquires more control rules. The Candide system [LZ94] has been used for the acquisition of a control knowledge. The system learns a qualitative model of the system first, then, based on the selected mode, it creates a rule based incremental controller for a given system. The SINS system [RS93] is a self improving reactive control system for autonomous robotic navigation. The system is based on using case based techniques and reinforcement learning. The learning module monitors the system and incrementally modifies the case representations to accommodate the changes. Because of the ....

A. Ram and J. C. Santamar'ia. Continuous case-based reasoning. In Proc. of the AAAI-93 Workshop on Case-Based Reasoning, Washington, DC, 1993.

.... or to understand the behaviour of an intelligent agent such as a human being, you have to analyse temporal sequences of interactions [18] Few existing works in Case Based Reasoning have tried to represent and use time extended situations inside cases in different applications: robot control [17], process forecast [19,16] process supervision [8] trend prognoses for medical problems [21] and medical risk detection and forecast [5] In these works, the representation and the retrieval of this kind of situations have shown many specificities compared to standard and instantaneous ....

....our framework for a plant nutrition control problem. 2. REQUIREMENTS OF OUR FRAMEWORK Our goal is to improve the management of past experiences based on our knowledge of different applications such as plant nutrition control (cf. 4) and on the analysis of the limitations of related works in CBR [17,19,16,8,21,5]. First, we will define the requirements to cope with a general class of problems in order to improve the reusability of the proposed management techniques. Then we will address four other issues that we want to integrate in order to extend the existing representation and retrieval methods. 2.1. ....

[Article contains additional citation context not shown here]

A. Ram and J.C. Santamaria. Continuous case-based reasoning. In AAAI Case-Based Reasoning Worksop, pages 86--93, 1993.

....set of indices giving the state of the world at a particular instant, and a time extended situation, a set of indices describing mainly the evolution of this state. Few existing applications in Case Based Reasoning have tried to represent and use behavioural situations inside cases: robot control [22], process forecast [23,21] process supervision [10] trend prognoses for medical problems [26] medical risk detection and forecast [6] plant nutrition control [11] WWW navigation [8,9] cf. 12] for a detailed analysis) To cope with this kind of situations, we have designed and implemented a ....

A. Ram and J.C. Santamaria. Continuous case-based reasoning. In AAAI Case-Based Reasoning Worksop, pages 86--93, 1993.

....regarding the implementation. We discuss further the conditions that are required for each approach to excel and support these arguments by presenting some comparison results on simulated experiments with the double integrator, pendulum swing up, and mobile robot systems (see also Atkeson and Santamar ia, 1997). We study in detail and in isolation each of the first three issues: continuous domains, imperfect perception, and incomplete knowledge in Chapters 4, 5, and 6, respectively. In each chapter we concentrate the research to the issue in hand while simplifying the other issues to facilitate the ....

....efficient updates, but requires high amount of data. The second procedure is based on value iteration (Bellman, 1957) it is a dynamic programming method that requires a model of the dynamic function, uses computational expensive updates, but requires less amount of data (Atkeson and Santamar ia, 1997). 3.2.4.1 Model Free Procedure Our model free procedure uses temporal difference to adapt the value function. Temporal difference updates adapt the weights of a function approximator based on the error or difference between two consecutive predictions. Sutton (1988) defines a whole family of ....

[Article contains additional citation context not shown here]

MA. Ram, A. and Santamar'ia, J. C. (1997). Continuous case-based reasoning. Artificial Intelligence, 90(1-2):25--77.

....a set of indices giving the state of the world at a particular instant, and a time extended situation, a set of indices describing mainly the evolution of this state. Few existing applications in CaseBased Reasoning have tried to represent and use behavioural situations inside cases: robot control [21], process forecast [22,20] process supervision [10] trend prognoses for medical problems [25] medical risk detection and forecast [5] plant nutrition control [11] WWW navigation [7,9] cf. 12] for a detailed analysis) To cope with this kind of situations, we have designed and implemented a ....

A. Ram and J.C. Santamaria. Continuous case-based reasoning. In AAAI Case-Based Reasoning Worksop, pages 86--93, 1993.

.... recognizes hand gestures (Darrell, Essa, Pentland 1996) for a simulated agent that maneuvers along a highway (McCallum 1996) and for a human computer interface that automates repetitive tasks (Das, Caglayan, Gonsalves 1998) Like Pierce and Kuipers (Pierce Kuipers 1997) Ram and Santamaria (Ram Santamaria 1997) and others, e.g. Iba 1991) we believe that sensorimotor agents can discover categories for themselves. Thus, the focus of this paper is an unsupervised method by which a mobile robot deduces meaningful categories from uninterpreted sensor readings. Previously, Rosenstein and Cohen demonstrated ....

....sensor readings. From Sensors to Categories For a mobile robot operating in an environment of even modest complexity, sensory categories supply a needed level of abstraction away from raw sensor readings (Mahadevan, Theocharous, Khaleeli 1998; Michaud Mataric 1998; Pierce Kuipers 1997; Ram Santamaria 1997). Since our objective is that agents discover such categories for themselves without supervision we adopt cluster analysis as a general, unsupervised technique that is well suited to the goals of this work. However, several subproblems exist, and this section outlines our solution to each ....

[Article contains additional citation context not shown here]

Ram, A., and Santamaria, J. C. 1997. Continuous case-based reasoning. Artificial Intelligence 90:25--77.

....used for nearest neighbor retrieval. Kelly and Davis (1991) use a genetic algorithm to find a vector of weightings of the attributes used in a nearest neighbor calculation in order to reduce the effects of irrelevant or misleading attributes and thus to make the distance measure more meaningful. Ram and Santamaria (1993) use continuous case based reasoning to perform tasks such as autonomous robotic navigation. They learn cases which provide information for the navigation system to deal with specific environments encountered. Our anytime learning system employs genetic algorithms to learn the most effective ....

Ram, A. and J. C. Santamaria (1993). Continuous Case-Based Reasoning. Case-Based Reasoning: Papers from the 1993 Workshop, Tech. Report WS-93-01, (pp 86-93). AAAI Press, Washington, D.C.

.... for a machine vision system that recognizes hand gestures (Darrell, Essa, Pentland 1996) for a simulated agent that maneuvers along a highway (McCallum 1996) and for a human computer interface that automates repetitive tasks (Das, Caglayan, Gonsalves 1998) Like Pierce and Kuipers (1997) Ram and Santamaria (1997) and others, e.g. Iba 1991; Thrun 1999) we believe that sensorimotor agents can discover categories for themselves. Thus, the focus of this paper is an unsupervised method by which a mobile robot deduces meaningful categories from uninterpreted sensor readings. Previously, we demonstrated a ....

....sensor readings. From Sensors to Categories For a mobile robot operating in an environment of even modest complexity, sensory categories supply a needed level of abstraction away from raw sensor readings (Mahadevan, Theocharous, Khaleeli 1998; Michaud Mataric 1998; Pierce Kuipers 1997; Ram Santamaria 1997). Since our objective is that agents discover such categories for themselves without supervision we make use of clustering techniques that offer a general, unsupervised framework for categorizing data. However, the following subproblems exist, and this section outlines our solution to each ....

[Article contains additional citation context not shown here]

Ram, A., and Santamaria, J. C. 1997. Continuous case-based reasoning. Artificial Intelligence 90:25--77.

....a new generation of expert system technology, relies on the retrieval, reuse, and revision of stored cases. Detailed descriptions of the technology can be found in books written by Kolodner [Kol93] and Leake [Lea96] Research has been conducted in various subfields of CBR; some references are [HH92, KM92, Sim92, SKR94, SC95, RS93]. In a case based reasoning system, each case is a combination of a problem description and a solution. The case base is organized to allow for efficient real time retrieval. Once retrieved, a user can adapt the case through simple modifications to solve their problem. In help desk applications, ....

A. Ram and J.C. Santamaria. Continuous case-based reasoning. Proceedings of the AAAI-93 Workshop on Case-Based REasoning, pages 86--93, 1993.

....Approximators Another class of sparse coarse coded memory is memory based. Although, memory based function approximators have not been widely used in conjunction with reinforcement learning, they are common in other tasks such as classification (e.g. 8] and robot control (e.g. 2] but see [16, 14, 11]) In a memory based function approximator, each memory element represents some of the state action pairs or a case the agent has experienced before. A query is performed by first retrieving the nearest neighbors to the query point according to some similarity metric and then performing a weighted ....

A. Ram and J. C. Santamar'ia. Continuous case-based reasoning. Artificial Intelligence, 90(12) :25--77, 1997.

....matrix and Equation 3 shows the complete hypothetical model for the this experiment. T = ff 0 Gamma ff EE (3) This model is a simplification of the model in equation 2 where the experience level (E) is the only regressor. Tables 9 and 10 show the statistical results for each individual parameter in the model as well as the 90 confidence interval estimation of its value for the learning and maturity phases, respectively. The results show that during the learning phase, the learning rate decreased from Gamma(fi E fi IE ) 4:89 seconds=experience to Gammaff E = 0:96 seconds=experience. This means that the system ....

Ram, A. & Santamar'ia, J.C. 1993b. Continuous Case-Based Reasoning, in Proceedings of the AAAI Workshop on Case-Based Reasoning, 86--93, AAAI Press.

No context found.

Intelligence. Ram, A. & Santamaria, J.C. (1997). Continuous CaseBased Reasoning. Artificial Intelligence, (90)1-2:25--77.

....Approximators Another class of sparse coarse coded memory is memory based. Although, memory based function approximators have not been widely used in conjunction with reinforcement learning, they are common in other tasks such as classification (e.g. 8] and robot control (e.g. 2] but see [16, 14, 11]) In a memory based memory, each memory element represents some of the state action pairs or a case the agent has experienced before. A query is performed by first retrieving the nearest neighbors to the query point according to some similarity metric and then performing a weighted average of ....

Ram, A. & Santama'ia, J.C. (1997). Continuous Case-Based Reasoning. To appear in Artificial Intelligence.

....(i.e. different sets of schema parameter values) as appropriate for the particular situation at hand. SINS can be characterized as performing a kind of constructive representational change in which it constructs higher level representations (cases) from low level sensorimotor representations (Ram, 1993). In SINS, the perception action task and the adaptation learning task are integrated in a tightly knit cycle, similar to the anytime learning approach of Grefenstette Ramsey (1992) Perception and action are required so that the system can explore its environment and detect regularities; ....

.... to traditional case based reasoning methods which perform high level reasoning in discrete, symbolic problem domains, SINS is based on a new method for continuous casebased reasoning in problem domains that involve continuous information, such as sensorimotor information for robot navigation (Ram Santamar a, 1993). There are still several unresolved issues in this research. The case retrieval process is very expensive and limits the number of cases that the system can handle without deteriorating the overall navigational performance, leading to a kind of utility problem (Minton, 1988) Our current solution ....

Ram, A. & Santamar a, J.C., Continuous CaseBased Reasoning, in Leake, D.B. (editor), Proceedings of the AAAI Workshop on Case-Based Reasoning, Washington, DC, 1993 (to appear).

No context found.

Ram, A. & Santamar'ia, J.C. 1993b. Continuous CaseBased Reasoning. In Proc. of the AAAI Workshop on Case-Based Reasoning, 86--93. Washington, DC.

No context found.

A. Ram and J.C. Santamar'ia, Continuous CaseBased Reasoning. In Leake, D.B. (editor), Proceedings of the AAAI Workshop on Case-Based Reasoning, AAAI Press Technical Report WS-9301, 86--93, Washington, DC, (1993).

....here) The latter condition guarantees termination since some worlds are unsolvable by one or more systems. Detailed experimental results may be found in Ram and Santamaria (1993a) Table 4 shows the final results after the execution of 200 randomly generated worlds, based on the design parameters max cases = 10, max cases = 4, and control interval = 12. SINS successfully navigates almost 95 of the worlds, a 555 improvement over the static system, with 43 fewer virtual collisions, 271 fewer steps, 38 shorter distance travelled, 36 better actual optimal ratio, and 40 speedup in performance time. ....

Ram, A. & Santamaria, J.C. (1993b). Continuous case-based reasoning. In Proceedings of the AAAI Workshop on Case-Based Reasoning, Washington, DC (to appear).

....of sparse coarse coded memory is memory based. Although, memory based function approximators have not been widely used in conjunction with reinforcement learning, they are common in other tasks such as classification (e.g. Kibler and Aha, 1989] and robot control (e.g. Atkeson, 1991] but see [Ram and Santamar ia, 1997; Peng, 1993; McCallum et al. 1995] In a memory based function approximator, each memory element represents some of the state action pairs or a case the agent has experienced before. A query is performed by first retrieving the nearest neighbors to the query point according to some similarity ....

Ram, A. and Santamar'ia, J. C. (1997). Continuous case-based reasoning. Artificial Intelligence, 90(1-2):25--77.

No context found.

A. Ram and J. C. Santamaria. Continuous case-based reasoning. Artificial Intelligence, 90(1-2):25--77, 1997.

No context found.

A. Ram and J.C. Santamaria, "Continuous casebased reasoning," Artificial Intelligence, 1998.

No context found.

A. Ram and J.C. Santamaria, "Continuous casebased reasoning," To appear in: Artificial Intelligence, 1998.

No context found.

A. Ram and J.C. Santamaria. Continuous case-based reasoning. Artificial Intelligence, 90:25--77, 1997.

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