Real-world applications of mobile robotics call for increased autonomy, requiring reliable perception systems. Since manually tuned perception algorithms are difficult to adapt to new operating environments, systems based on supervised learning are necessary for future progress in autonomous navigation. Data labeling is a major concern when supervised learning is applied to the large-scale problems occuring in realistic robotics applications. We believe that algorithms for automatically selecting important data for labeling are necessary, and propose to employ active learning techniques to reduce the amount of labeling required to learn from a data set. In this paper we show that several standard active learning algorithms can be adapted to meet specific constraints characteristic to our domain, such as the need to learn from data with severely unbalanced class priors. We validate the solutions we propose by extensive experimentation on multiple realistic data sets captured with a robotic vehicle. Based on our results for the task of obstacle detection, we conclude that active learning techniques are applicable to our domain, and they can lead to significant reductions in the labeling effort required to use supervised learning in outdoor perception.
{"title":"Active Learning For Outdoor Obstacle Detection","authors":"C. Dima, M. Hebert","doi":"10.15607/RSS.2005.I.002","DOIUrl":"https://doi.org/10.15607/RSS.2005.I.002","url":null,"abstract":"Real-world applications of mobile robotics call for increased autonomy, requiring reliable perception systems. Since manually tuned perception algorithms are difficult to adapt to new operating environments, systems based on supervised learning are necessary for future progress in autonomous navigation. Data labeling is a major concern when supervised learning is applied to the large-scale problems occuring in realistic robotics applications. We believe that algorithms for automatically selecting important data for labeling are necessary, and propose to employ active learning techniques to reduce the amount of labeling required to learn from a data set. In this paper we show that several standard active learning algorithms can be adapted to meet specific constraints characteristic to our domain, such as the need to learn from data with severely unbalanced class priors. We validate the solutions we propose by extensive experimentation on multiple realistic data sets captured with a robotic vehicle. Based on our results for the task of obstacle detection, we conclude that active learning techniques are applicable to our domain, and they can lead to significant reductions in the labeling effort required to use supervised learning in outdoor perception.","PeriodicalId":87357,"journal":{"name":"Robotics science and systems : online proceedings","volume":"61 1","pages":"9-16"},"PeriodicalIF":0.0,"publicationDate":"2005-06-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"80757679","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper presents an integrated approach to exploration, mapping, and localization. Our algorithm uses a highly efficient Rao-Blackwellized particle filter to represent the posterior about maps and poses. It applies a decision-theoretic framework which simultaneously considers the uncertainty in the map and in the pose of the vehicle to evaluate potential actions. Thereby, it trades off the cost of executing an action with the expected information gain and takes into account possible sensor measurements gathered along the path taken by the robot. We furthermore describe how to utilize the properties of the Rao-Blackwellization to efficiently compute the expected information gain. We present experimental results obtained in the real world and in simulation to demonstrate the effectiveness of our approach.
{"title":"Information Gain-based Exploration Using Rao-Blackwellized Particle Filters","authors":"C. Stachniss, G. Grisetti, Wolfram Burgard","doi":"10.15607/RSS.2005.I.009","DOIUrl":"https://doi.org/10.15607/RSS.2005.I.009","url":null,"abstract":"This paper presents an integrated approach to exploration, mapping, and localization. Our algorithm uses a highly efficient Rao-Blackwellized particle filter to represent the posterior about maps and poses. It applies a decision-theoretic framework which simultaneously considers the uncertainty in the map and in the pose of the vehicle to evaluate potential actions. Thereby, it trades off the cost of executing an action with the expected information gain and takes into account possible sensor measurements gathered along the path taken by the robot. We furthermore describe how to utilize the properties of the Rao-Blackwellization to efficiently compute the expected information gain. We present experimental results obtained in the real world and in simulation to demonstrate the effectiveness of our approach.","PeriodicalId":87357,"journal":{"name":"Robotics science and systems : online proceedings","volume":"52 3 1","pages":"65-72"},"PeriodicalIF":0.0,"publicationDate":"2005-06-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"91055811","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper demonstrates that sonar images correlate badly because target geometry and image fringes correlate in different coordinate systems. It is shown that if the receiver is a line array, all point objects have the same image in a coordinate system with axes of range and the sine of target bearing (the (r, s) coordinate system). Results from an ocean experiment are presented. The ocean experiment shows that after a simple translation, the correlation coefficient between Cartesian images of a radar reflector drop to zero, while the correlation coefficient between images of the same radar reflector in an (r, s) coordinate system hover around 0.9.
{"title":"On Correlating Sonar Images","authors":"Richard J. Rikoski, J. T. Cobb, Daniel C. Brown","doi":"10.15607/RSS.2005.I.023","DOIUrl":"https://doi.org/10.15607/RSS.2005.I.023","url":null,"abstract":"This paper demonstrates that sonar images correlate badly because target geometry and image fringes correlate in different coordinate systems. It is shown that if the receiver is a line array, all point objects have the same image in a coordinate system with axes of range and the sine of target bearing (the (r, s) coordinate system). Results from an ocean experiment are presented. The ocean experiment shows that after a simple translation, the correlation coefficient between Cartesian images of a radar reflector drop to zero, while the correlation coefficient between images of the same radar reflector in an (r, s) coordinate system hover around 0.9.","PeriodicalId":87357,"journal":{"name":"Robotics science and systems : online proceedings","volume":"87 1","pages":"169-176"},"PeriodicalIF":0.0,"publicationDate":"2005-06-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81935935","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
A. Gueorguiev, P. Allen, Ting Song, A. Laine, W. Edstrom, John E. Hunt
We present a microrobotic system for protein crystal micromanipulation tasks. The focus in this paper is on the task known to crystallographers as streak seeding – it is used to entice certain protein crystals to grow. Our system features a set of custom designed micropositioner end-effectors we call microshovels to replace traditional tools used by crystallographers for this task. We have used micro-electrical mechanical system (MEMS) techniques to design and manufacture various shapes and quantities of microshovels. Visual input from a camera mounted on the microscope is used to detect the locations of the source crystals which the tool needs to touch as well as the locations of the target protein droplets for seeding. We present experimental results that illustrate the applicability of our approach.
{"title":"Microrobotic Streak Seeding For Protein Crystal Growth","authors":"A. Gueorguiev, P. Allen, Ting Song, A. Laine, W. Edstrom, John E. Hunt","doi":"10.15607/RSS.2005.I.019","DOIUrl":"https://doi.org/10.15607/RSS.2005.I.019","url":null,"abstract":"We present a microrobotic system for protein crystal micromanipulation tasks. The focus in this paper is on the task known to crystallographers as streak seeding – it is used to entice certain protein crystals to grow. Our system features a set of custom designed micropositioner end-effectors we call microshovels to replace traditional tools used by crystallographers for this task. We have used micro-electrical mechanical system (MEMS) techniques to design and manufacture various shapes and quantities of microshovels. Visual input from a camera mounted on the microscope is used to detect the locations of the source crystals which the tool needs to touch as well as the locations of the target protein droplets for seeding. We present experimental results that illustrate the applicability of our approach.","PeriodicalId":87357,"journal":{"name":"Robotics science and systems : online proceedings","volume":"65 1","pages":"137-144"},"PeriodicalIF":0.0,"publicationDate":"2005-06-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83968159","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper presents Monte-Carlo localization (MCL) [1] with a mixture proposal distribution for mobile robots with stereo vision. We combine filtering with the Scale Invariant Feature Transform (SIFT) image descriptor to accurately and efficiently estimate the robot’s location given a map of 3D point landmarks. Our approach completely decouples the motion model from the robot’s mechanics and is general enough to solve for the unconstrained 6-degrees of freedom camera motion. We call our approach σMCL. Compared to other MCL approaches σMCL is more accurate, without requiring that the robot move large distances and make many measurements. More importantly our approach is not limited to robots constrained to planar motion. Its strength is derived from its robust vision-based motion and observation models. σMCL is general, robust, efficient and accurate, utilizing the best of Bayesian filtering, invariant image features and multiple view geometry techniques.
{"title":"σMCL: Monte-Carlo Localization for Mobile Robots with Stereo Vision","authors":"P. Elinas, J. Little","doi":"10.15607/RSS.2005.I.049","DOIUrl":"https://doi.org/10.15607/RSS.2005.I.049","url":null,"abstract":"This paper presents Monte-Carlo localization (MCL) [1] with a mixture proposal distribution for mobile robots with stereo vision. We combine filtering with the Scale Invariant Feature Transform (SIFT) image descriptor to accurately and efficiently estimate the robot’s location given a map of 3D point landmarks. Our approach completely decouples the motion model from the robot’s mechanics and is general enough to solve for the unconstrained 6-degrees of freedom camera motion. We call our approach σMCL. Compared to other MCL approaches σMCL is more accurate, without requiring that the robot move large distances and make many measurements. More importantly our approach is not limited to robots constrained to planar motion. Its strength is derived from its robust vision-based motion and observation models. σMCL is general, robust, efficient and accurate, utilizing the best of Bayesian filtering, invariant image features and multiple view geometry techniques.","PeriodicalId":87357,"journal":{"name":"Robotics science and systems : online proceedings","volume":"23 1","pages":"373-380"},"PeriodicalIF":0.0,"publicationDate":"2005-06-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"86985249","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
A large group of robots will often be partitioned into subgroups, each subgroup performing a different task. This paper presents four distributed algorithms for assigning swarms of homogenous robots to subgroups to meet a specified global task distribution. Algorithm Random-Choice selects tasks randomly, but runs in constant time. Algorithm Extreme-Comm compiles a complete inventory of all the robots on every robot, runs quickly, but uses a great deal of communication. The CardDealer’s algorithm assigns tasks to individual robots sequentially, using minimal communications but a great deal of time. The TreeRecolor algorithm is a compromise between Extreme-Comm and Card-Dealer’s, balancing communications use and running time. The three deterministic algorithms drive the system towards the desired assignment of subtasks with high accuracy. We implement the algorithms on a group of 25 iRobot SwarmBots, and collect and analyze performance data.
{"title":"Dynamic Task Assignment in Robot Swarms","authors":"J. McLurkin, Daniel Yamins","doi":"10.15607/RSS.2005.I.018","DOIUrl":"https://doi.org/10.15607/RSS.2005.I.018","url":null,"abstract":"A large group of robots will often be partitioned into subgroups, each subgroup performing a different task. This paper presents four distributed algorithms for assigning swarms of homogenous robots to subgroups to meet a specified global task distribution. Algorithm Random-Choice selects tasks randomly, but runs in constant time. Algorithm Extreme-Comm compiles a complete inventory of all the robots on every robot, runs quickly, but uses a great deal of communication. The CardDealer’s algorithm assigns tasks to individual robots sequentially, using minimal communications but a great deal of time. The TreeRecolor algorithm is a compromise between Extreme-Comm and Card-Dealer’s, balancing communications use and running time. The three deterministic algorithms drive the system towards the desired assignment of subtasks with high accuracy. We implement the algorithms on a group of 25 iRobot SwarmBots, and collect and analyze performance data.","PeriodicalId":87357,"journal":{"name":"Robotics science and systems : online proceedings","volume":"20 1","pages":"129-136"},"PeriodicalIF":0.0,"publicationDate":"2005-06-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73238645","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Future space missions are expected to use robotic systems to assemble, inspect, and maintain large space structures in orbit. For effective planning and control, robots must know the deformations and motions of the structures with which they interact. This paper presents a method for estimating the shape, motion, and dynamic model parameters of a vibrating space structure using asynchronous raster-scanning range imagers. The method assumes that the mode shapes are approximately known a priori. A Kalman filter exploits a mechanics-based dynamic model to extract the modal frequencies and damping as well as the modal coefficients and their time rate of change. Theoretical development and experimental results using emulated space hardware are presented. Index Terms – space structure, laser rangefinder, shape estimation, motion estimation, cooperative sensing.
{"title":"Shape, Motion, and Parameter Estimation of Flexible Space Structures using Laser Rangefinders","authors":"M. Lichter, S. Dubowsky, H. Ueno, S. Mitani","doi":"10.15607/RSS.2005.I.046","DOIUrl":"https://doi.org/10.15607/RSS.2005.I.046","url":null,"abstract":"Future space missions are expected to use robotic systems to assemble, inspect, and maintain large space structures in orbit. For effective planning and control, robots must know the deformations and motions of the structures with which they interact. This paper presents a method for estimating the shape, motion, and dynamic model parameters of a vibrating space structure using asynchronous raster-scanning range imagers. The method assumes that the mode shapes are approximately known a priori. A Kalman filter exploits a mechanics-based dynamic model to extract the modal frequencies and damping as well as the modal coefficients and their time rate of change. Theoretical development and experimental results using emulated space hardware are presented. Index Terms – space structure, laser rangefinder, shape estimation, motion estimation, cooperative sensing.","PeriodicalId":87357,"journal":{"name":"Robotics science and systems : online proceedings","volume":"63 1","pages":"351-358"},"PeriodicalIF":0.0,"publicationDate":"2005-06-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83049299","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
P. Abbeel, Adam Coates, Michael Montemerlo, A. Ng, S. Thrun
Kalman filters are a workhorse of robotics and are routinely used in state-estimation problems. However, their performance critically depends on a large number of modeling parameters which can be very difficult to obtain, and are often set via significant manual tweaking and at a great cost of engineering time. In this paper, we propose a method for automatically learning the noise parameters of a Kalman filter. We also demonstrate on a commercial wheeled rover that our Kalman filter’s learned noise covariance parameters—obtained quickly and fully automatically—significantly outperform an earlier, carefully and laboriously hand-designed one.
{"title":"Discriminative Training of Kalman Filters","authors":"P. Abbeel, Adam Coates, Michael Montemerlo, A. Ng, S. Thrun","doi":"10.15607/RSS.2005.I.038","DOIUrl":"https://doi.org/10.15607/RSS.2005.I.038","url":null,"abstract":"Kalman filters are a workhorse of robotics and are routinely used in state-estimation problems. However, their performance critically depends on a large number of modeling parameters which can be very difficult to obtain, and are often set via significant manual tweaking and at a great cost of engineering time. In this paper, we propose a method for automatically learning the noise parameters of a Kalman filter. We also demonstrate on a commercial wheeled rover that our Kalman filter’s learned noise covariance parameters—obtained quickly and fully automatically—significantly outperform an earlier, carefully and laboriously hand-designed one.","PeriodicalId":87357,"journal":{"name":"Robotics science and systems : online proceedings","volume":"15 1","pages":"289-296"},"PeriodicalIF":0.0,"publicationDate":"2005-06-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"78690537","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
M. Lagoudakis, E. Markakis, D. Kempe, P. Keskinocak, A. Kleywegt, Sven Koenig, C. Tovey, A. Meyerson, Sonal Jain
Recently, auction methods have been investigated as effective, decentralized methods for multi-robot coordination. Experimental research has shown great potential, but has not been complemented yet by theoretical analysis. In this paper we contribute a theoretical analysis of the performance of auction methods for multi-robot routing. We suggest a generic framework for auction-based multi-robot routing and analyze a variety of bidding rules for different team objectives. This is the first time that auction methods are shown to offer theoretical guarantees for such a variety of bidding rules and team objectives.
{"title":"Auction-Based Multi-Robot Routing","authors":"M. Lagoudakis, E. Markakis, D. Kempe, P. Keskinocak, A. Kleywegt, Sven Koenig, C. Tovey, A. Meyerson, Sonal Jain","doi":"10.15607/RSS.2005.I.045","DOIUrl":"https://doi.org/10.15607/RSS.2005.I.045","url":null,"abstract":"Recently, auction methods have been investigated as effective, decentralized methods for multi-robot coordination. Experimental research has shown great potential, but has not been complemented yet by theoretical analysis. In this paper we contribute a theoretical analysis of the performance of auction methods for multi-robot routing. We suggest a generic framework for auction-based multi-robot routing and analyze a variety of bidding rules for different team objectives. This is the first time that auction methods are shown to offer theoretical guarantees for such a variety of bidding rules and team objectives.","PeriodicalId":87357,"journal":{"name":"Robotics science and systems : online proceedings","volume":"23 1","pages":"343-350"},"PeriodicalIF":0.0,"publicationDate":"2005-06-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"77943417","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
We consider the problem of planning manipulation tasks in which rigid body dynamics are significant and the rigid bodies undergo frictional contacts. We develop a dynamic model with frictional compliant contacts, and a time-stepping algorithm that lends itself to finding trajectories with constraints on the starting and goal conditions. Because we explicitly model the local compliance at the contact points, we can incorporate impacts without resetting the states and reinitializing the dynamic models. The problem of solving for the frictional forces with the Coulomb friction cone law reduces to a convex quadratic program. We show how this formulation can be used to solve boundary value problems that are relevant to process design, design optimization and trajectory planning with practical examples. To our knowledge, this paper is the first time boundary value problems involving changes in contact conditions have been solved in a systematic way.
{"title":"A Two-Point Boundary-Value Approach for Planning Manipulation Tasks","authors":"Peng Song, Vijay Kumar, J. Pang","doi":"10.15607/RSS.2005.I.017","DOIUrl":"https://doi.org/10.15607/RSS.2005.I.017","url":null,"abstract":"We consider the problem of planning manipulation tasks in which rigid body dynamics are significant and the rigid bodies undergo frictional contacts. We develop a dynamic model with frictional compliant contacts, and a time-stepping algorithm that lends itself to finding trajectories with constraints on the starting and goal conditions. Because we explicitly model the local compliance at the contact points, we can incorporate impacts without resetting the states and reinitializing the dynamic models. The problem of solving for the frictional forces with the Coulomb friction cone law reduces to a convex quadratic program. We show how this formulation can be used to solve boundary value problems that are relevant to process design, design optimization and trajectory planning with practical examples. To our knowledge, this paper is the first time boundary value problems involving changes in contact conditions have been solved in a systematic way.","PeriodicalId":87357,"journal":{"name":"Robotics science and systems : online proceedings","volume":"59 1","pages":"121-128"},"PeriodicalIF":0.0,"publicationDate":"2005-06-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"85825553","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：