Notice
2.7. Grid Localization: an example in 1D
- document 1 document 2 document 3
- niveau 1 niveau 2 niveau 3
Descriptif
Now that we have theequations of the Bayes filter, we need a method in order to implementin real cases these equations.So, in the following, I want todiscuss two methods, which are commonly adopted by the MobileRobotics Community and, these, if you want, correspond to twoextreme solutions because one is a fully numerical and itis based on a grid and, for the case of localization, isknown as the grid-localization approach – and theother one is a fully analytical and it isknown as a the Kalman filter.So, now, in this video, wediscuss the first method, which is numerical, and we do thisby referring to a simple case – a trivial case, aone-dimensional case – and this will allow not only to understand thesemethods – how it works – but also to better understandthe behavior of the Bayes filter.
Thème
Documentation
Liens
Dans la même collection
-
2.1. Localization process in a probabilistic framework: basic concepts
MartinelliAgostinoIn this part, we will talk about localization which is a fundamental problem that a robot has to be able to solve in order to accomplish almost any tasks. In particular, we will start by
-
2.6. The Bayes Filter
MartinelliAgostinoThe equations of the Bayes filters are the equation that allow us to update the probability distribution for the robot to be in a given configuration by integrating the information that are in the
-
2.3. Wheel encoders for a differential drive vehicle
MartinelliAgostinoIn this video, we want to discuss the case of a wheel encoders in 2D, and in particular the case of a robot equipped with a differential drive which is very popular in mobile robotics.
-
2.4. Sensor statistical models
MartinelliAgostinoSo far in the characterization of our sensor measurements, we didn't talk about the errors. This is precisely what we want to do in this video. In particular, we want to compute two probability
-
2.8. The Extended Kalman Filter (EKF)
MartinelliAgostinoWe have seen the grid localization, and the advantage of this approach is that we can deal with any kind of probability distribution; in particular we don't need to do a Gaussian assumption. The
-
2.2. Characterization of proprioceptive and exteroceptive sensors
Before deriving the equations of the Bayes filter, I want to remind you a few concepts in the theory of probability, and also some mathematical characterization for the statistical error of the
-
2.5. Reminds on probability
MartinelliAgostinoIn this sequence I want to remind you a few concepts in the theory of probability and then in the next one we finally derive the equations of the Bayes filter. So the concept that I want to
Avec les mêmes intervenants et intervenantes
-
3.8. Applications of the Observability Rank Criterion
MartinelliAgostinoIn this video we want to apply the observability rank criterion to understand the observability properties of the system that we saw in the previous videos.
-
3.6. Observability in robotics
MartinelliAgostinoIn this video we discuss a fundamental issue which arises when we deal with an estimation problem: understanding if the system contains enough information to perform the estimation of the state.
-
3.1. Examples for the Action in the EKF
MartinelliAgostinoIn part 2, we have seen the equations of the Bayes filter, which are the general equations which allow us to update the probability distribution, as the data from both proprioceptive sensors and
-
3.4. The use of the EKF in robotics
MartinelliAgostinoIn this video I want to explain the steps that we have to follow in order to implement an extended Kalman filter in robotics.
-
3.2. Examples for the Perception in the EKF
MartinelliAgostinoIn this video we discuss the second two equations of the Kalman filter.
-
3.5. Simultaneous Localization and Mapping (SLAM)
MartinelliAgostinoIn this video, we are discussing the SLAM problem: simultaneous localization and mapping.
-
3.7. Observability Rank Criterion
MartinelliAgostinoIn this video, we discuss an automatic method which is analytical and allows us to answer the question if a state is observable or not: this method is the Observability Rank Criterion which has
-
3.3. The EKF is a weight mean
MartinelliAgostinoIn this video I want to discuss the second two equations of the Kalman filter. And in particular I want to show that these actually perform a kind of weight mean.
-
2.4. Sensor statistical models
MartinelliAgostinoSo far in the characterization of our sensor measurements, we didn't talk about the errors. This is precisely what we want to do in this video. In particular, we want to compute two probability
-
2.6. The Bayes Filter
MartinelliAgostinoThe equations of the Bayes filters are the equation that allow us to update the probability distribution for the robot to be in a given configuration by integrating the information that are in the
-
2.1. Localization process in a probabilistic framework: basic concepts
MartinelliAgostinoIn this part, we will talk about localization which is a fundamental problem that a robot has to be able to solve in order to accomplish almost any tasks. In particular, we will start by
-
2.5. Reminds on probability
MartinelliAgostinoIn this sequence I want to remind you a few concepts in the theory of probability and then in the next one we finally derive the equations of the Bayes filter. So the concept that I want to
Sur le même thème
-
Et si l’intelligence artificielle déferlait sur les océans ?
L’évolution des technologies d’observation et de modélisation a joué un rôle central dans l’accroissement des connaissances sur le fonctionnement des océans ou dans le développement des activités
-
La vidéo sous-marine au service de la recherche halieutique
Accessible à de nombreuses applications, tant en biologie ou qu'en technologie des pêches, la vidéo sous-marine est de plus en plus utilisée dans le domaine de la recherche halieutique. Les progrès
-
5.8. Other approaches: Social Forces
Vasquez GoveaAlejandro DizanIn this video we will review one of the alternatives we are proposing to the use of Hidden Markov models and typical trajectories: the Social Force model.
-
5.6. Predicting Human Motion
Vasquez GoveaAlejandro DizanIn video 5.5 we have defined an HMM in Python. In this video we are going to learn how to use it to estimate and predict motion.
-
5.9. Other approaches: Planning-based approaches
Vasquez GoveaAlejandro DizanIn this video we are going to study a second, and probably the most promising alternative for motion prediction: planning-based algorithms.
-
5.4. Bayesian filter inference
Vasquez GoveaAlejandro DizanIn this video we will review the base filter and we will study a particular instance of the Bayesian filter called Hidden Markov models which is a discrete version of a Bayesian filter.
-
5.7. Typical Trajectories: drawbacks
Vasquez GoveaAlejandro DizanIn previous videos we have discussed how to implement the typical trajectories and motion patterns approach. In this video we are going to discuss what are the drawbacks of such an approach,
-
5.5. From trajectories to discrete time-state models
Vasquez GoveaAlejandro DizanIn this video we are going to apply the concepts we have reviewed in the video 5.4 into real trajectories.
-
4.3. Dynamic Probabilistic Grids – Implementation approaches
LaugierChristianThis video addresses the problem of the practical implementation for the dynamic probabilistic grid.
-
5.1. Introduction
Vasquez GoveaAlejandro DizanIn this part of the course we are going to go deeper into situation awareness and in particular we are going to study how to model and learn human behavior which is a crucial task for social-aware
-
4.5. Detection and Tracking of Mobile Objects – Problem and Approaches
LaugierChristianThis video adresses the Detection and Tracking of Mobile Objects (DATMO) problem.
-
5.3b. Learning typical trajectories 2/2
Vasquez GoveaAlejandro DizanIn this video we are aiming to improve on the results we obtained in video 5.3a, in particular with respect to the greyed-out trajectories that are badly represented.
