The project focuses on developing Collaboration Simultaneous Localisation and Mapping solution that allows accurate pose and map estimation in 3D environments. The computationally efficient collaboration logic is intended for deployment in multiple robots with low computational resources such as quadcopters.
The single-robot 3D SLAM has been developed in such a way that a UGV is able to propagate through a single global feature map based on the information collected from a 3D depth sensor, an IMU and a stereo camera. In CSLAM for multiple UGV, the information of all sensors (depth sensor, IMU and stereo camera) are fused at each time step by the centralised fusion algorithm and the overall results are sent back to each robot. As part of the project package, a novel autonomous 3D exploration algori
thm has been developed for two UGVs that guides the robotic system with a tremendous reduction in computational effort.
Overview of single-robot 3D SLAM Overview of 3D CSLAM
3D mapping generated using UGV with autonomous 3D exploration
- Yue. Yuefeng., Wang, Danwei., Senarathne, P.G.C.N., Moratuwage, M.D.P. " A Hybrid Probabilistic and Point Set Registration Approach for Fusion of 3D Occupancy Grid Maps." 2016 IEEE International Conference on Systems, Man and Cybernetics (SMC 2016).
- Senarathne, P.G.C.N, Wang Danwei. "Towards autonomous 3D exploration using surface frontiers." 2016 IEEE International Symposium on Safety, Security and Rescue Robotics (SSRR 2016)