W. Burgard, D. Fox, and D. Hennig. Fast grid-based position tracking for mobile robots. In G. Brewka, C. Habel, and B. Nebel, editors, Proceedings of the 21st Annual German Conference on Artificial Intelligence (KI-97): Advances in Artificial Intelligence, pages 289--300. Springer, 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....the sensor impressions. Due to the lack of a closed expression for the distribution function, it has to be approximated. One appropriate model is provided by grid based Markov localization approaches that have been examined for some time: they either use sonar sensors [4] or laser range finders [2] to create a probability grid. As a result, a hypothesis about the current position of the robot can be inferred from that grid. Recently, so called Monte Carlolocalization approaches have become very popular. They use particle filters to approximate the distribution function [7, 25] As a ....

....maps (one for each node of the topological map) 5. 2 Comparison between RouteLoc and prominent approaches A number of prominent self localization algorithms use the Markov localization approach, some of them with toplogical representations of the environment [23, 15, 27] others with metric maps [2, 7, 25]. In the robotics community, it is referred to as Markov localization if the algorithm somehow exploits the socalled Markov assumption [22] It states that the outcome of a state transition may only depend on the current state and the chosen action. The outcome does explicitly not depend on ....

W. Burgard, D. Fox, and D. Henning. Fast grid-based position tracking for mobile robots. In G. Brewka, Ch. Habel, and B. Nebel, editors, KI-97: Advances in Artificial Intelligence, Lecture Notes in Artificial Intelligence, pages 289--300, Berlin, Heidelberg, New York, 1997. Springer.

....for example, assume a large number of evenly spaced sensors, which render them useless in robots with very few sensors. In comparison to mobile robots that have a ring of sonars, the sensing capabilities of the Pioneer 1 are rather limited. At rst, we tried to use a Markov localisation method [2]. However, this approach failed. The robot became lost when the sonar sensor readings were sparse and noisy, for example, when the robot was moving diagonally through a corridor. In this situation, the walls of the corridor re ect the sonar beams and hardly any distance readings from the front ....

W. Burgard, D. Fox, and D. Hennig. Fast grid-based position tracking for mobile robots. In G. Brewka, C. Habel, and B. Nebel, editors, Proceedings of the 21st Annual German Conference on Arti cial Intelligence (KI-97): Advances in Arti cial Intelligence, pages 289-300. Springer, 1997.

....capapble of handling uncertain and ambiguous information. Among the possible forms forms of representation, we can distinguish between (1) geometric representations, such as hypotheses about landmark features [6] and (2) grid based representations, such as position probability grids [4]. It seems that both representations have their justi cation. The localisation problem can be divided into two subproblems: 1) the estimation of the absolute position in the environment, usually referred to as absolute localisation, and (2) the tracking of the robot s position relative to a ....

....a simple robot with seven sensors. In our experimental setup, the robot had to collect objects in an oce like environment. The pose of the robot was required as input to a learning algorithm which controlled the robot motors and the gripper. At rst, we tried to use a Markov localisation method [4]. However, this approach failed. The robot became lost when the sonar sensor readings were sparse and noisy, for example, when the robot was moving diagonally through a corridor. The Pioneer 1 s seven ultrasonic proximity sensors are one on each side and ve forward facing. In comparison to mobile ....

[Article contains additional citation context not shown here]

W. Burgard, D. Fox, and D. Hennig. Fast grid-based position tracking for mobile robots. In G. Brewka, C. Habel, and B. Nebel, editors, Proceedings of the 21st Annual German Conference on Articial Intelligence (KI-97): Advances in Articial Intelligence, pages 289-300. Springer, 1997.

.... s motion. Thus, there is a risk of overlooking obstacles when using such a static fire strategy, not only for the Bremen Autonomous Wheelchair, but also for other mobile robots with sonar sensors, e.g. the Nomad 200 manufactured by Nomadic (used e.g. in [13] or RWI s B21 (used e.g. in [4]) Section 3.3 will present a dynamic fire strategy that solves this problem. To further reduce the impact of cross talks on the wheelchair s behavior, only obstacles are taken into account that were measured at least twice. The dynamic fire strategy that will be presented in section 3.3 ensures ....

W. Burgard, D. Fox, and D. Henning. Fast grid-based position tracking for mobile robots. In G. Brewka, Ch. Habel, and B. Nebel, editors, KI-97: Advances in Artificial Intelligence, Lecture Notes in Artificial Intelligence, pages 289--300, Berlin, Heidelberg, New York, 1997. Springer.

....of landmark features, such as corners and wall segments. Markov localisation and scan matching form the third category. The majority of existing localisation methods are passive, that is, independent of the robot control. Recently, also active approaches to localisation have been proposed [ Burgard et al. 1997a ] Active localisation methods are able to change the robot motion and the orientation of the sensors in order to increase the e ciency and the robustness of localisation. The robot s position and orientation, also referred to as pose, have to be determined from sensor data. Unfortunately, ....

....capable of handling uncertain and ambiguous information. Among the possible forms of representation, we can distinguish between geometric representations, such as hypotheses about landmark features [ Drumheller, 1987 ] and grid based representations, such as position probability grids [ Burgard et al. 1997b ] Both representations have interesting and useful features. The localisation problem can be divided into two subproblems: 1) the estimation of the absolute position in the environment, usually referred to as absolute localisation, and (2) the tracking of the robot s position relative to a ....

[Article contains additional citation context not shown here]

Wolfgang Burgard, Dieter Fox, and Daniel Hennig. Fast grid-based position tracking for mobile robots. In Gerhard Brewka, Christopher Habel, and Bernhard Nebel, editors, Proceedings of the 21st Annual German Conference on Articial Intelligence (KI-97): Advances in Articial Intelligence, pages 289-300. Springer, 1997.

....extension would be to apply the proposed method to other types representations, with different sensor models. The perceptual model used here, which is based on landmarks, is just one choice out of many possible choices. A different choice would be the probabilistic sensor model described in (Burgard, Fox, Hennig, 1997; Burgard et al. 1996) which specifically applies to proximity sensors, such as sonars or laser range finders. The inverse sensor model (also called sensor interpretation) which is employed in the map building step (M step) can be realized by the approach described in (Thrun, 1998) where ....

Burgard, W., Fox, D., & Hennig, D. (1997). Fast grid-based position tracking for mobile robots. Proceedings of the 21th German Conference on Artificial Intelligence (pp. 289--300). Berlin: Springer Verlag.

No context found.

Wolfram Burgard, Dieter Fox, and Daniel Hennig. Fast grid-based position tracking for mobile robots. In Proc. of the 21th German Conference on Artificial Intelligence (KI'97), Freiburg, Germany, 1997. Springer Verlag.

No context found.

W. Burgard, D. Fox, and D. Hennig. Fast grid-based position tracking for mobile robots. In Proc. of the 21th German Conference on Artificial Intelligence (KI 97), Freiburg, Germany. Springer Verlag, 1997. to appear.

No context found.

W. Burgard, D. Fox, and D. Hennig. Fast grid-based position tracking for mobile robots. In Proc. of the 21th German Conference on Artificial Intelligence, Germany. Springer Verlag, 1997.

....our current implementation of Markov localization lies in the xed discretization of the state space, which is always kept in main memory. To scale up to truly large environments, it seems inevitable that one needs variable resolution representations of the state space, suchasastheone suggested in (Burgard et al. 1997; 1998b; Gutmann et al. 1998) Alternatively, one could use Monte Carlo based representations of the state space as described in (Fox et al. 1999) Here, the robot s belief is represented by samples that concentrate on the most likely parts of the state space. Acknowledgment The authors ....

W. Burgard, D. Fox, and D. Hennig. Fast grid-based position tracking for mobile robots. In Proc. of the German ConferenceonArticial Intelligence (KI), Germany. Springer Verlag, 1997.

....current implementation of Markov localization lies in the xed discretization of the state space, which is always kept in main memory. To scale up to truly large environments, it seems inevitable that one needs variable resolution representations of the state space, such as as the one suggested in (Burgard et al. 1997; 1998b; Gutmann et al. 1998) Alternatively, one could use Monte Carlo based representations of the state space as described in (Fox et al. 1999) Here, the robot s belief is represented by samples that concentrate on the most likely parts of the state space. Acknowledgment The authors ....

W. Burgard, D. Fox, and D. Hennig. Fast grid-based position tracking for mobile robots. In Proc. of the German Conference on Articial Intelligence (KI), Germany. Springer Verlag, 1997.

.... state space: l = P l2R(lm ) l p(L t = l) P l2R(lm ) p(L t = l) 3) In contrast to the topological approaches described [11, 14, 8] which use predefined landmarks to compute p(s j l) we obtain this likelihood directly from a metric model of the environment and a model of proximity sensors [4]. The advantage of this approach is that it can operate based on the raw data of the proximity sensors and thus permits the exploitation of arbitrary geometric features of the environment such as the width of a corridor or the size of a cupboard. However, it can easily be extended to incorporate ....

W. Burgard, D. Fox, and D. Hennig. Fast grid-based position tracking for mobile robots. In Proceedings of the 21th German Conference on Artificial Intelligence (KI 97), Freiburg, Germany. Springer Verlag, 1997.

....of 7; 200; 000 states. To deal with such huge state spaces in real time, which is essential for fast position position estimation, we apply different optimization techniques. First, we use a fast sensor model allowing to compute the sensing probability P (s j l) by simple look up operations [BFH97] The second optimization approach is a technique for a selective update of the grid. The key idea of this approach is to exclude unlikely cells in P from being updated. For this purpose, we introduce a threshold 1 and approximate P (s j l) for cells with P (l) by P (s) which is the ....

W. Burgard, D. Fox, and D. Hennig. Fast grid-based position tracking for mobile robots. In Proc. of the 21th German Conference on Artificial Intelligence, Germany. Springer Verlag, 1997.

....in our current implementation stays constant even if only a minor part of the state space is updated. In this context we would like to mention that recently promising techniques have been presented to overcome this disadvantage by applying alternative and dynamic representations of the state space [6,5,12]. 3.2 The Model of Proximity Sensors As mentioned above, the likelihood P (s j l) that a sensor reading s is measured at position l has to be computed for all positions l in each update cycle. Therefore, it is crucial for on line position estimation that this quantity can be computed very ....

....of future research. First, the current implementation of Markov localization uses a fixed discretization of the whole state space which is always kept in memory. To overcome this disadvantage, recently different alternative representations of the density have been developed and suggested [6,5,15]. Whereas [6] uses a local grid for efficient position tracking, we introduced an Octree based representation of the robot s state space in [5] which allows to dynamically adopt the required memory as well as the resolution of the discretization. 15] suggest a combination of Markov localization ....

[Article contains additional citation context not shown here]

W. Burgard, D. Fox, and D. Hennig. Fast grid-based position tracking for mobile robots. In Proc. of the 21st German Conference on Artificial Intelligence, Germany. Springer Verlag, 1997.

....is the probability of measuring s at location l. In DWA, the sensors are assumed to measure proximity and s are proximity measurements (obtained from laser range finders and or sonar sensors) P (s j l) is obtained using the map and a simplistic sensor model, which is described in more detail in [4]. When the robot moves, P is convolved using a probabilistic model of robot motion: P (l) Gamma X l 0 P (l j u; l 0 ) P (l 0 ) 2) where P (l j u; l 0 ) denotes the probability that the robot is at l upon executing control u at position l 0 . In DWA, P (l j u; l 0 ) is ....

....i ) 4) Figure 5 depicts the density of the measurement X ff in the two situations shown in Figures 3 and 4: one, in which the robot 1 In fact, in our implementation d ff (l) is computed in advance for all possible l and ff and stored in a look up table, which maximizes run time efficiency. See [4] for a more detailed discussion of efficient retrieval. does not know its position well, and one where it is fairly certain about its position. As can easily be seen, when the robot is uncertain about its position, X ff is spread over many different measurements (solid line) If the robot knows ....

W. Burgard, D. Fox, and D. Hennig. Fast grid-based position tracking for mobile robots. In Proc. of the 21th German Conference on Artificial Intelligence (KI 97), Germany. Springer Verlag, 1997.

....ultrasound sensors and laser range finders. Please note that the higher accuracy of laserrange finders versus ultrasound sensors is represented by a smaller standard deviation of the Gaussian distribution. With this sensor model, computing p(sjl) amounts to a fast series of two table look ups (see [5] for more details) 2) Selective Computation. Most of the time the probability mass is centered on a small number of location. With the exception of the initial global localization phase, the vast majority of probabilities Bel(L = l) are usually close to 0 and can safely be ignored. This ....

W. Burgard, D. Fox, and D. Hennig. Fast grid-based position tracking for mobile robots. In Proc. of the 21th German Conference on Artificial Intelligence, Germany. Springer Verlag, 1997.

No context found.

W. Burgard, D. Fox, and D. Hennig. Fast grid-based position tracking for mobile robots. In G. Brewka, C. Habel, and B. Nebel, editors, Proceedings of the 21st Annual German Conference on Artificial Intelligence (KI-97): Advances in Artificial Intelligence, pages 289--300. Springer, 1997.

No context found.

Wolfram Burgard, Dieter Fox, and Daniel Hennig. Fast Grid-based Position Tracking for Mobile Robots. In Proc. of the 21th German Conference on Artificial Intelligence, volume 1303 of LNCS. Springer-Verlag, 1997.

No context found.

W. Burgard, D. Fox, and D. Hennig. Fast grid-based position tracking for mobile robots. In Proc. of the German Conference on Artificial Intelligence (KI), number 1303 in Lecture Notes in Computer Science, Germany, 1997. Springer-Verlag.

No context found.

Wolfram Burgard, Dieter Fox, and Daniel Hennig. Fast Grid-based Position Tracking for Mobile Robots. In Proc. of the 21th German Conference on Artificial Intelligence, volume 1303 of LNCS. Springer-Verlag, 1997.

No context found.

W. Burgard, D. Fox, and D. Hennig. Fast grid-based position tracking for mobile robots. In G. Brewka, C. Habel, and B. Nebel, editors, Proceedings of the 21st Annual German Conference on Artificial Intelligence (KI-97): Advances in Artificial Intelligence, pages 289--300. Springer, 1997.

No context found.

W. Burgard, D. Fox, and D. Hennig. Fast grid-based position tracking for mobile robots. In G. Brewka, C. Habel, and B. Nebel, editors, Proceedings of the 21st Annual German Conference on Articial Intelligence (KI-97): Advances in Articial Intelligence, pages 289--300. Springer, 1997.

No context found.

W. Burgard, D. Fox, and D. Henning, "Fast grid-based position tracking for mobile robots," in KI-97: Advances in Artificial Intelligence, G. Brewka, Ch. Habel, and B. Nebel, Eds., Berlin, Heidelberg, New York, 1997, Lecture Notes in Artificial Intelligence, pp. 289--300, Springer.

No context found.

Wolfram Burgard, Dieter Fox, and Daniel Hennig. Fast gridbased position tracking for mobile robots. In Proceedings of the 21th German Conference on Artificial Intelligence (KI 97), Freiburg, Germany. Springer Verlag, 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