Results for outdoor-SLAM using sparse extended information filters
read more
Citations
The Graph SLAM Algorithm with Applications to Large-Scale Mapping of Urban Structures
Simultaneous Localization and Mapping: A Survey of Current Trends in Autonomous Driving
Simultaneous Localization and Mapping
Hierarchical SLAM: real-time accurate mapping of large environments
Simultaneous Localization and Map Building in Large-Scale Cyclic Environments Using the Atlas Framework
References
A solution to the simultaneous localization and map building (SLAM) problem
On the representation and estimation of spatial uncertainty. [for mobile robot]
On the representation and estimation of spatial uncertainly
Estimating Uncertain Spatial Relationships in Robotics
Related Papers (5)
Frequently Asked Questions (13)
Q2. What is the difficult problem in SLAM?
In their simulations, the authors focused particularly on the loop closing problem, which is generally acknowledged to be one of the hardest problems in SLAM.
Q3. How can the authors implement data association in logarithmic time?
Using kd-trees, it appears to be feasible to implement data association in logarithmic time by recursively partitioning the space of all landmark locations using a tree.
Q4. What is the mechanism for handling the data association problem?
Their mechanism for handling the data association problem uses a maximum likelihood estimator, together with a thresholded χ2 test.
Q5. What is the common approach for calculating p(nt|zt, ?
In EKFs, calculating p(nt|zt, ut) is easy since it is straightforward to extract the mean and the covariance of a landmark position together with the robot pose from the full state estimate.
Q6. What is the gradient of the measurement function h with respect to the state vector?
Ct is the gradient of the measurement function h with respect to the state vector ξ, taken at ξ = µt:Ct = ∇ξh(µt) (8)In general filter applications, such an update may require more than constant time.
Q7. Why does SEIF have bigger error than EKF?
Due to the approximation of the information matrix and amortized map recovery, SEIF has bigger error than EKF as is shown in Figure 12.
Q8. How do you recover t in a linear algorithm?
SEIFs use an amortized iterative method, similar to the Jacobi method or the slightly different GaussSeidel method, to gradually recover µt.
Q9. What is the effect of removing links between Y 0 and the robot pose?
Their sparsification equations have the effect of removing links between Y 0 and the robot pose — a step necessary if the number of landmarks linked to the robot exceeds a given sparsity threshold.
Q10. What is the effect of the motion on the robot?
This reflects the fact that even though the motion induces a loss of information of the robot relative to the landmarks, no information is lost between individual landmarks.
Q11. What is the main purpose of this paper?
This paper summarized a new algorithm for the simultaneous localization and mapping (SLAM) problem, which can maintain globally consistent maps with constant update time.
Q12. How is the update rule used to maintain sparseness?
Sparseness is achieved by an update rule that occasionally removes links from the posterior so as to maintain sparseness, as described further below.
Q13. How is the odometer used for the vehicle?
The raw odometry of the vehicle is extremely poor, resulting in several hundred meters of error when used for path integration along the vehicle’s 3.5km path, see Figure 8(a).