The angle between the landmark and the platform the sensor is mounted on, changes with the movement of the platform, resulting in varying numbers of radar detections over time
Clutter in the scene may result in false detections
At each time step the radar measurements are sifted, the selected measurements which are within a pre-defined distance of the existing landmarks are used for measurement association and to update the state.
Measurements which are not selected are used to determine if there are any new landmarks.
To determine whether an existing landmark should be deleted or an unconfirmed landmark should be confirmed we use M/N logic - if a landmark has been observed M times in the last N timesteps it is confirmed/deleted.
graph TD
A(k = 1) --> B(EKF Prediction)
B --> C(Get radar measurements Zk)
C --> S1(Calculate distances landmarks to measurements)
S1 --> S2{Distances < threshold?}
S2 --"Yes"--> G(Associate measurements)
S2 --"No"--> H(Cluster measurements)
G --> I(Delete landmarks)
H --> J(Confirm landmarks)
J --> K(Merge landmarks)
I --> D(k = k + 1)
K --> D
D --> B
classDef highlight fill:#9cd9eb,stroke:#f92080,stroke-width:1px;
class G,H,I,J,K highlight;
Landmark management scheme, highlighted blocks are explained below.
The first step is to check whether the selected measurements can be associated with existing landmarks based on a log distance measure.
The associated measurements are used to update the EKF state.
Landmark Deletion
graph LR
D1(For each landmark in state)
D1 --> D2{Landmark associated with measurements?}
D2 --"Yes"--> D3(Update landmark)
D2 --"No"--> D4{Landmark within range?}
D4 --"Yes"--> D5{Landmark meets M/N logic criteria?}
D5 --"Yes"--> D6(Delete landmark)
Landmark deletion procedure.
In this step if a landmark is not associated with any measurements and is close enough to the sensor that it should be observed, we use M/N logic determine if the landmark should be deleted.
Deleting landmarks allows us to remove any which may have been incorrectly initialized or may have left the scene (i.e. a car has now left the car park).
Measurements which were not initially selected as close to an existing landmark are clustered.
If a cluster is within a specified distance of an existing landmark it is used to update the landmark otherwise it is considered an unconfirmed landmark.
Measurements are sifted with each red diamond representing a landmark, remaining measurements are clustered, with measurements that do not belong to a cluster treated as false detections.
If the number of measurements associated with the unconfirmed landmark are above a defined threshold
If the landmark has been observed often enough according to the M/N logic
Once a landmark is confirmed it is added to the system state.
Landmark confirmation rules.
Landmark Merging
graph LR
M1(Calculate distances between landmarks) --> M2{Distances < threshold?}
M2 --"Yes"--> M3(Merge landmarks)
Finally, there is still a possibility that two landmarks in the system may in reality relate to the same landmark.
If the distance between two landmarks is less than a set threshold they are merged.
Example EKF-SLAM with Landmark Management
Example of landmark management using EKF SLAM
Extended Targets
Current work [2] is looking at how to produce a more precise map, in order to better facilitate navigation and path planning.
To do this we are using an ellipse model to generate extended landmarks and investigating how to take advantage of the landmark extent estimation in the update of the EKF-SLAM.
This paper focuses on efficient landmark management in radar based simultaneous localization and mapping (SLAM). Landmark management is necessary in order to maintain a consistent map of the estimated landmarks relative to the estimate of the platform’s pose. This task is particularly important when faced with multiple detections from the same landmark and/or dynamic environments where the location of a landmark can change. A further challenge with radar data is the presence of false detections. Accordingly, we propose a simple yet efficient rule based solution for radar SLAM landmark management. Assuming a low-dynamic environment, there are several steps in our solution: new landmarks need to be detected and included, false landmarks need to be identified and removed, and the consistency of the landmarks registered in the map needs to be maintained. To illustrate our solution, we run an extended Kalman filter SLAM algorithm in an environment containing both stationary and temporally stationary landmarks. Our simulation results demonstrate that the proposed solution is capable of reliably managing landmarks even when faced with false detections and multiple detections from the same landmark.
Mapping Extended Landmarks for Radar SLAM
Shuai Sun, Christopher Gilliam, Kamran Ghorbani, Glenn Matthews, and Beth Jelfs
Simultaneous localization and mapping (SLAM) using automotive radar sensors can provide enhanced sensing capabilities for autonomous systems. In SLAM applications, with a greater requirement for the environment map, information on the extent of landmarks is vital for precise navigation and path planning. Although object extent estimation has been successfully applied in target tracking, its adaption to SLAM remains unaddressed due to the additional uncertainty of the sensor platform, bias in the odometer reading, as well as the measurement non-linearity. In this paper, we propose to incorporate the Bayesian random matrix approach to estimate the extent of landmarks in radar SLAM. We describe the details for implementation of landmark extent initialization, prediction and update. To validate the performance of our proposed approach we compare with the model-free ellipse fitting algorithm with results showing more consistent extent estimation. We also demonstrate that exploiting the landmark extent in the state update can improve localization accuracy.
