Citation Context

...akness lies in the strong assumptions that have to be made on both the robot motion model and the sensor noise. Moreover, the landmarks are assumed to be uniquely identifiable. There exist techniques =-=[33]-=- to deal with unknown data association in the SLAM context, however, if these assumptions are violated, the filter is likely to diverge [12]. Similar observations have been reported by Julier et al. [...

Citation Context

...uces the total computational burden to O(n 2 log n). The approach we have adopted for Map-Matching is not unique. For example the Joint Compatibility test with branch and bound technique suggested by =-=[13]-=- could also be utilized. The freedom to choose a Map-Matching strategy highlights the modularity of the Atlas framework. E. Traversal Once the robot has mapped an area, it can reuse previously built m...

Citation Context

...ization process. As will be proved in Section III-A, this allows us to maintain the computational cost of loop closing linear with the size of the loop. C. Building Local Maps During continuous SLAM, local maps are sequentially built in the following way (see [8] for a more detailed explanation of the process): at a given instant , a new map is initialized using the current vehicle location as base reference . Then, the vehicle continues moving and acquiring sensor information about the environment features . At each step, data association is robustly solved using the joint compatibility test [22], and the local map is updated with standard stochastic mapping techniques [23]. The resulting local map is independent of any prior map because it is built relative to the initial vehicle location at and depends only on the sequence of odometry readings and sensor data obtained during the involved steps. Under the common assumption that process and measurement noise are white random sequences, two local maps built with the same robot from disjoint sequences of steps are functions of independent stochastic variables. Therefore, the two maps will be statistically independent and uncorrelated. T...

Citation Context

.... New landmarks are acquired during motion (from [14]). Data Association: The standard formulation of the EKF-SLAM solution is especially fragile to incorrect association of observations to landmarks =-=[35]-=-. The ‘loop-closure’ problem, when a robot returns to re-observe landmarks after a large traverse, is especially difficult. The association problem is compounded in environments where landmarks are no...

Citation Context

...at remains to be discussed is precisely how the occurrence of a loop closure can be detected. A simple and sometimes effective approach would be to look to the SLAM p.d.f to detect loop closure as in =-=[16]-=-. A sufficiently small Mahalanobis distance between two poses could then indicate loop closure. However, this assumes that the estimate of the underlying joint p.d.f. is not in gross error. Unfortunat...

Citation Context

...lding processes. The initial uncertainty in B21a’s position was set to zero. That of B21b however was set to tens of meters. The joint compatibility branch and bound (JCBB) data association algorith=-=m [33]-=- was used by both vehicles and facilitated the initialization of B21b into the map being built by B21a. The initial large uncertainty of B21b meant that the data association step implicitly localized ...