Scene Modelling, Recognition and Tracking with Invariant Image Features
ISMAR '04 Proceedings of the 3rd IEEE/ACM International Symposium on Mixed and Augmented Reality
Consistent, convergent, and constant-time SLAM
IJCAI'03 Proceedings of the 18th international joint conference on Artificial intelligence
CI-graph: an efficient approach for large scale SLAM
ICRA'09 Proceedings of the 2009 IEEE international conference on Robotics and Automation
CI-graph: an efficient approach for large scale SLAM
ICRA'09 Proceedings of the 2009 IEEE international conference on Robotics and Automation
SURF: speeded up robust features
ECCV'06 Proceedings of the 9th European conference on Computer Vision - Volume Part I
Sparse Local Submap Joining Filter for Building Large-Scale Maps
IEEE Transactions on Robotics
Hi-index | 0.00 |
In this paper, a graph-based hierarchical SLAM framework is proposed which ensures not only the high-speed operation of graph-based SLAM, but also feasibility of large-scale 3D map building. Both local and global level graph-based SLAM will be operated. They can be concurrently implemented in different computing units but must maintained communication which is called as a session. During each session, local level SLAM will create a pose-graph containing trajectory for mobile robot and information for local maps. To avoid communication congestion, raw density point cloud of each pose-node will be reduced into a sparser one by voxel grid filtering. When a session is closed, duo-graph strategy is executed to guarantee the consistency between successive local map. To associate massive local map information and closing loops in large-scale environment, graph-based algorithm will also be implemented in global level SLAM. Two experiments are carried out in the real indoor environment. In the first experiment, SLAM process is greatly accelerated in this framework by distributing the whole SLAM task into different level SLAM entities: local level SLAM operated on robot and global on laptop. It shows that this framework is scalable and it can be implemented in CS(Client-Server) model. Second experiment is conducted in our biggest work office around which we control the robot traverse for three times. It demonstrates that this framework can simultaneously keep fast graph-based SLAM in local-end and generate consistent and convergent 3D map in global-end SLAM process.