I have a ground mobile robot with wheel for indoor applications and I want to develop a vSLAM algorithm for it. So one problem for me is how to know my modifications do really improve the performance of vSLAM. Since I want to test it in real home environment, it is not feasible to install a motion capture system. So my questions are:
- Could you recommend some public vslam datasets for indoor ground robots?
- Can consumer level 1d lidar provides ground truth? such as rpLidar? Since the environment is not that friendly. What I concern is that I am not sure if lidar can provide much more accurate pose as ground truth in the place where vSLAM can't meet my accuracy target, such as mirror, windows/door with transparent glasses.
- Are there any methods to evaluate vSLAM accuracy without ground truth?
I really appreciate your answers. Thanks.