R. Greiner and R. Isukapalli. Learning to select useful landmarks. IEEE Transactions on Systems, Man, and Cybernetics, Part B:473--449, 1996.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....a reluctant work for us to arrange artificial landmarks. Visual features such as vertical edges of door or obstacle are used as landmarks[2, 3] However, such features may not be stable, that is, may not be observable under various conditions of lighting and background scene changed by viewpoints[4]. Therefore, it is desirable to determine stable landmarks based on observed data. Proc. of IROS 99, pp. 781 786, Kyongju, Korea, Oct. 1999. Many researcher used learning and execution strategies with human assistance in order to select landmarks from observed data[5, 6] The robot tries to ....

R. Greiner and R. Isukapalli "Learning to Select Useful Landmarks," IEEE Trans. on Systems, Man, and Cybernetics-PartB:Cybernetics ,Vol. 26, No. 3, pp. 437-449, 1996.

.... itself [36] Dark bright regions and vertical edges are used in [13, 71] and hallways, openings and doors are used by the approaches described in [34, 61, 62] Others have proposed methods for learning what feature to extract, through a training phase in which the robot is told its location [27, 54, 65]. These are just a few representative examples of many different features used for localization. Our approach differs from all these approaches in that it does not extract predefined features from the sensor values. Instead, it directly processes raw sensor data. Such an approach has two key ....

R. Greiner and R. Isukapalli. Learning to select useful landmarks. In Proc. of the National Conference on Artificial Intelligence (AAAI), pages 1251--1256, Menlo Park, CA, 1994. AAAI Press / The MIT Press.

....is greater than the other with high con dence; we need not wait for the sample size speci ed by the Cherno bound, which we have to when the frequencies are similar. Sequential sampling methods have been reported to reduce the required sample size by several orders of magnitude (e.g. [10]) In our algorithm (Table 1) we combine sequential sampling with the popular loop reversal technique found in many KDD algorithms. Instead of processing hypotheses one after another, and obtaining enough examples for each hypothesis to evaluate it suciently precisely, we keep obtaining ....

R. Greiner and R. Isukapalli. Learning to select useful landmarks. IEEE Transactions on Systems, Man, and Cybernetics, Part B:473-449, 1996.

....will yield the best localization result. Sutherland and Thompson [11] developed one of the earliest methods for landmark selection. They applied heuristic functions to select a landmark triple, from the set of such triples, that is likely to yield a good localization result. Greiner and Isukapalli [2] learn a function to select landmarks that minimize the expected localization error. A related technique is given by Thrun [13] who trains a neural network to learn landmarks that optimize the localization uncertainty. Yeh and Kriegman [14] select the subset of features from a set of possible ....

R. Greiner and R. Isukapalli. Learning to select useful landmarks. IEEE Transactions on Systems, Man, and Cybernetics - Part B : Cybernetics, 26(3):437-449, June 1996.

.... itself [36] Dark bright regions and vertical edges are used in [13, 71] and hallways, openings and doors are used by the approaches described in [34, 61, 62] Others have proposed methods for learning what feature to extract, through a training phase in which the robot is told its location [27, 54, 65]. These are just a few representative examples of many different features used for localization. Our approach differs from all these approaches in that it does not extract predefined features from the sensor values. Instead, it directly processes raw sensor data. Such an approach has two key ....

R. Greiner and R. Isukapalli. Learning to select useful landmarks. In Proc. of the National Conference on Artificial Intelligence (AAAI), pages 1251--1256, Menlo Park, CA, 1994. AAAI Press / The MIT Press.

....is greater than the other with high con dence; we need not wait for the sample size speci ed by the Cherno bound, which we have to when the frequencies are similar. Sequential sampling methods have been reported to reduce the required sample size by several orders of magnitude (e.g. [7]) In our algorithm (Table 1) we combine sequential sampling with the popular loop reversal technique found in many KDD algorithms. Instead of processing hypotheses one after another, and obtaining enough examples for each hypothesis to evaluate it suciently precisely, we keep obtaining ....

....learning context by proposing the Hoe ding Race algorithm that combines loop reversal with adaptive Hoe ding bounds. A general scheme for sequential local search has been proposed by Greiner [8] Sequential sampling can often reduce the required sample sizes in cases by considerable factors [7]. Sampling techniques are particularly needed in the context of knowledge discovery in databases where often much more data are available than can be processed. A non sequential sampling algorithm for KDD has been presented by Toivonen [17] a sequential algorithm (that imposes further ....

R. Greiner and R. Isukapalli. Learning to select useful landmarks. IEEE Transactions on Systems, Man, and Cybernetics, Part B:473-449, 1996.

....a reluctant work for us to arrange artificial landmarks. Visual features such as vertical edges of door or obstacle are used as landmarks[2, 3] However, such features may not be stable, that is, may not be observable under various conditions of lighting and background scene changed by viewpoints[4]. Therefore, it is desirable to determine stable landmarks based on observed data. 3 Proc. of IROS 99, pp. 781 786, Kyongju, Korea, Oct. 1999. Many researcher used learning and execution strategies with human assistance in order to select landmarks from observed data[5, 6] The robot tries to ....

R. Greiner and R. Isukapalli "Learning to Select Useful Landmarks," IEEE Trans. on Systems, Man, and Cybernetics-PartB:Cybernetics , Vol. 26, No. 3, pp. 437-449, 1996.

.... itself [80] Dark bright regions and vertical edges are used in [31,159] and hallways, openings and doors are used by the approach described in [82,135,138] Others have proposed methods for learning what feature to extract, through a training phase in which the robot is told its location [58,114,145,146]. These are just a few representative examples of many different features used for localization. RHINO s approach differs from all these approaches in that it does not extract predefined features from the sensor values. Instead, it directly processes raw sensor data. Such an approach has two key ....

R. Greiner and R. Isukapalli. Learning to select useful landmarks. In Proceedings of 1994 AAAI Conference, pages 1251--1256, Menlo Park, CA, 1994. AAAI Press / The MIT Press.

.... itself [39] Dark bright regions and vertical edges are used in [13, 74] and hallways, openings and doors are used by the approach described in [41, 62, 63] Others have proposed methods for learning what feature to extract, through a training phase in which the robot it told its location [28, 56, 67, 68]. These are just a few representative examples of many different features used for localization. Our approach differs from all these approaches in that it does not extract predefined features from the sensor values. Instead, it directly processes raw sensor data. Such an approach has two key ....

R. Greiner and R. Isukapalli. Learning to select useful landmarks. In Proceedings of 1994 AAAI Conference, pages 1251--1256, Menlo Park, CA, 1994. AAAI Press / The MIT Press.

....is manifold, given the enormous number of different search techniques which could be used. A large class of those techniques systematically use knowledge generated within the system and supported by outside methods, e.g. statistics (for a review, see Kaelbling et al. 1996) or PAC learning (Greiner Isukapalli, 1996). Algorithms inspired by natural processes, on the other hand, are applicable as well. Many connectionist approaches have been proposed to achieve learning for autonomous agents robots. Mill an in (Mill an, 1996) has worked on a neural network architecture for the acquisition of efficient ....

Greiner, R., & Isukapalli, R. (1996). Learning to Select Useful Landmarks. In IEEE Transactions Systems, Man and Cybernetics - Part B, Special Issue on Learning Autonomous Robots, 26, 437 --- 449.

....compute the robot position and heading relative to a 2D floor map. The simplicity of this approach, and the fact that it does not involve any 3D reconstruction, has made it popular. At least three landmarks are required in order to compute the position, but more can also be used as for example in [8]. This paper studies the use of exactly three landmarks (a landmark triplet) for triangulation. The focus is on how accurately the robot position can be estimated using this approach. The sensory input to the triangulation method is the location in the image of three landmarks. The output is robot ....

R. Greiner and R. Isukapalli. Learning to select useful landmarks. IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics, 26(3):437 -- 449, June 1996. Special issue on learning in autonomous robots.

.... itself [70] Dark bright regions and vertical edges are used in [26, 135] and hallways, openings and doors are used by the approach described in [72, 116, 119] Others have proposed methods for learning what feature to extract, through a training phase in which the robot it told its location [49, 101, 125, 126]. These are just a few representative examples of many different features used for localization. 38 Experiences with an Interactive Museum Tour Guide Robot RHINO s approach differs from all these approaches in that it does not extract predefined features from the sensor values. Instead, it ....

R. Greiner and R. Isukapalli. Learning to select useful landmarks. In Proceedings of 1994 AAAI Conference, pages 1251--1256, Menlo Park, CA, 1994. AAAI Press / The MIT Press.

....different contexts, and found that the palo 1 system discussed here was usually the best, in terms of the utility of its final performance element, as a function of the empirical sample complexity. More recently, however, we found that palo 1N worked effectively in one particular context; see [35]. 6.2 Limitations The examples discussed in Section 5 illustrate the versatility and generality of the palo objective, of identifying a performance element whose expected utility, over an arbitrary (but stationary) distribution of problems, is optimal. Our particular palo system is designed to ....

R. Greiner and R. Isukapalli. Learning to select useful landmarks. IEEE Transactions on Systems, Man and Cybernetics, accepted subject to modifications.

....by sketching three meaningful applications, which provide concrete solutions to the utility problem from explanation based learning, the multiple extension problem from non monotonic reasoning and the tractability completeness tradeoff problem from knowledge representation. The subsequent papers [14, 15] show that a robot can use this same general idea, and algorithm, on the very different task of learning the best set of landmarks to use for registering its location. These papers also provide a large corpus of experiments that demonstrate that Palo works very effectively in this context as ....

Russell Greiner and Ramana Isukapalli. Learning to select useful landmarks. IEEE Transactions on Systems, Man and Cybernetics, Part B, 26(3), June 1996. ($GreinerFTP/useful-lms-smc.ps).

....by sketching three meaningful applications, which provide concrete solutions to the utility problem from explanation based learning, the multiple extension problem from non monotonic reasoning and the tractability completeness tradeoff problem from knowledge representation. The subsequent papers [14, 15] show that a robot can use this same general idea, and algorithm, on the very different task of learning the best set of landmarks to use for registering its location. These papers also provide a large corpus of experiments that demonstrate that Palo works very effectively in this context as ....

Russell Greiner and Ramana Isukapalli. Learning to select useful landmarks. In Proceedings of AAAI-94, 1994. ($GreinerFTP/useful-lms-aaai.ps).

....different contexts, and found that the palo 1 system discussed here was usually the best, in terms of the utility of its final performance element, as a function of the empirical sample complexity. More recently, however, we found that palo 1N worked effectively in one particular context; see [35]. 6.2 Limitations The examples discussed in Section 5 illustrate the versatility and generality of the palo objective, of identifying a performance element whose expected utility, over an arbitrary (but stationary) distribution of problems, is optimal. Our particular palo system is designed to ....

R. Greiner and R. Isukapalli. Learning to select useful landmarks. IEEE Transactions on Systems, Man and Cybernetics, accepted subject to modifications.

No context found.

R. Greiner and R. Isukapalli. Learning to select useful landmarks. IEEE Transactions on Systems, Man, and Cybernetics, Part B:473--449, 1996.

No context found.

R. Greiner and R. Isukapalli. Learning to select useful landmarks. IEEE Transactions on Systems, Man, and Cybernetics, Part B:473-449, 1996.

No context found.

Russell, Greiner, and Isukapalli. Learning to select useful landmarks. IEEE Transactions on Systems, Man and Cybernetics, Part B, 26(3), 1996.

No context found.

B. Greiner and R. Isukapalli. Learning to Select Useful Landmarks. IEEE Trans. on Systems, Man, and Cybernetics - Part B: Cybernetics, 26(3):437--449, 1996.

No context found.

Russell, Greiner, and Isukapalli, "Learning to select useful landmarks," IEEE Transactions on Systems, Man and Cybernetics, Part B, vol. 26, no. 3, 1996. [Online]. Available: citeseer.nj.nec.com/article/greiner94learning.html

No context found.

Greiner, R., & Isukapalli, R. (1994). Learning to select useful landmarks. Proceedings of National Conference on Artificial Intelligence (pp. 1251--1256). Menlo Park, CA: AAAI Press / The MIT Press.

No context found.

R. Greiner, R. Isukapalli, Learning to select useful landmarks, in: Proc. AAAI-94, Seatle, WA, AAAI Press/MIT Press, Menlo Park, CA, 1994, pp. 1251--1256.

No context found.

R. Greiner and R. Isukapalli, "Learning to select useful landmarks," IEEE Transactions on Systems, Man, and Cybernetics-Part B, Special Issue on Robot Learning, 26, 437-449, 1996.

No context found.

Greiner, R., & Isukapalli, R. (in press). Learning to select useful landmarks. IEEE Transactions on Systems, Man, and Cybernetics . Konolige, K. (1995). A refined method for occupancy grid interpretation. Proceedings of the International Workshop on Uncertainty in Robotics . Amsterdam, Netherlands.

First 50 documents

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