About CV, SLAM, robot navigation, broken-down reading

Contact slam for about half a year, this time, we follow up the latest open source algorithm, and according to the actual situation modified some algorithm source code, realize the robot slam, path planning and obstacle avoidance behavior. And I also from a distance from the door 18 Street small white smooth into the door.

Before due to weak foundation, the computer vision, SLAM, robot navigation some concepts are not very clear, resulting in the selection of the scheme and the sensor, the general direction is a little vague, did not jump out of the local view of the global, took some detours. With the deepening of learning, some basic knowledge also has a relatively clear concept, recently tried to comb the relevant things, but also a summary of their own.

First a block diagram of my own freehand drawing:

1, previously believed that the robot navigation map should be the slam output map, but in the graph optimization era, the visual slam output of the map, it seems to be only used to track the camera trajectory. After acquiring the camera pose, the robot needs to use other sensors, such as LIDAR, to construct the obstacle map needed for the navigation. That is, the maps used for navigation and visual slam output do not have to be confused.

2, Slam is a branch of computer vision, if you want to use visual to do some other applications, such as some visual interaction and recognition, because the visual calculation of the consumption of resources, in the system hardware architecture should not be done with slam, itself should be two different research directions and application scenarios.

3, CV, slam, robot navigation, there will be some cross, but essentially belong to different research direction, each have their own methods and difficulties, if the team want to do more in-depth, should be three parts of research work by different people to bear. Division I now a little confusion, basically everyone involved in these three directions, so efficiency will be relatively low.

The above is my own comb things, comparative basis, and some are also obvious things, if there are errors in the place, hope to see friends can be generous enlighten. The more you learn the more you feel your ignorance, the more things you need to learn.

About CV, SLAM, robot navigation, broken-down reading