Abstract
With the advent of 3D sensors, point clouds are becoming increasingly popular in robotic perception. Using point clouds, mapping algorithms can generate 3D environment models. A widely used one is OctoMap, informing a robotic platform which parts of the environment are free and which are not using the octree structure.The generation of point clouds from sensor data and the operation of OctoMap are governed by different parameters, the correct selection of which significantly affects the process and the quality of the final map. Unfortunately, research in the process to identify the parameter set to achieve best occupancy mapping performance remains limited. The current work aims to fill this gap with a two-step principled methodology that first identifies the most significant parameters by conducting Neighbourhood Component Analysis on all parameters and then optimise those using grid search with the area under the Receiver Operating Characteristic curve.
In addition, the map update policy in OctoMap has limitations. All the nodes containing endpoints will be assigned with the same probability regardless of the points being noise and the probability of one such node can only be increased with a single measurement. Moreover, potentially occupied nodes with points inside but traversed by rays cast from the sensor to endpoints will be marked as free. To overcome these limitations, the current work presents a mapping method using the context of neighbouring points to update the nodes containing points, with the occupancy information of a point represented by the average distance from the point to its k-Nearest Neighbours (k-NN). A relationship between the distance and the change in probability is defined with the Cumulative Density Function of average distances, potentially decreasing the probability of a node despite points being present inside.
This study is conducted on 20 data sets collected with specially designed targets in two outdoor environments, providing precise ground truths for evaluation purposes. Point clouds are created by applying StereoSGBM on the images from a stereo camera and poses are produced by ORB-SLAM. Using the proposed method, a clear indication can be seen that the mapping parameters are more important than other parameters. Through grid search, improvement in occupancy map quality can be achieved over default parameters. Moreover, the k-NN mapping method can also achieve improvement over the performance of OctoMap.
| Date of Award | 18 Jul 2022 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Ioannis Georgilas (Supervisor), Alan Hunter (Supervisor) & Andrew Plummer (Supervisor) |