The secret of autonomous movement of robots: What is the relationship between slam and path planning? Three

Blog reprinted from: https://www.leiphone.com/news/201612/lvDXqY82OGNqEiyl.html

Lei Feng Net (public number: Lei Feng net) Press: This article author Slamtec (think LAN technology SLAMTEC-SH) technical consultant, focus on slam and related sensor research and development applications.

Let's first look at the relationship between slam and path planning.

In fact, the SLAM algorithm itself only completes the localization of the robot and the map building Two things, and we say that the navigation is not exactly equivalent to the positioning. The navigation here, in fact, is slam algorithm can not do. It is called Motion planning in the industry (motion planning).

Motion planning is a very large concept, from the movement of the robot arm, to the flight of the aircraft, and then here we say sweeping path planning, is the category of sports planning.

Let's talk about the motion planning of this kind of wheeled robot for sweeping machine. The basic capability required here is path planning, which is the ability to navigate a target point after completing the slam. In layman's terms, plan a path from point A to point B and let the robot move past.

To achieve this, motion planning implements at least two levels of modules, one called Global planning , and this is a little bit like our car navigator, which needs to plan a line ahead of the map, as well as the position of the current robot. This is provided by our slam system. The industry will generally use the algorithm called A * to achieve this process, it is a heuristic search algorithm, very good. Its most application is in the game, such as StarCraft, Warcraft and other real-time strategy games, are using this algorithm to calculate the motion trajectory of the unit.

Of course, just planning the path is not enough, there will be a lot of unexpected situation in reality, such as just happened to have a child in the way, you need to adjust the original path. Of course, sometimes this adjustment does not need to recalculate the global path, the robot may be a little bit around a bend. At this point, we need another level of planning module, called Local planning . It may not know where the robot will eventually go, but it is particularly adept at how the robot can bypass the obstacles in front of it.

Video please see the source blog

These two levels of planning module work together, the robot can be very good to achieve from point A to point B action, but the actual work environment, the above configuration is not enough. For example, a * algorithm planning path is based on a known map, pre-planned, once the robot to the destination of the process encountered a new obstacle, it had to stop completely, waiting for obstacles to leave or re-planning the path. If the sweeping robot buys home, it must first walk the house again before the floor, the user experience will be very poor.

For this reason, there will also be improvements to such algorithms, such as Slamware we use the improved d* algorithm for path planning, which is the core pathfinding algorithm used by the American Mars rover . This is a dynamic heuristic path search algorithm that allows robots to move freely in unfamiliar environments and navigate the changing environment.

The biggest advantage of the d* algorithm is that it does not require a pre-proven map, and the robot can act like a human, even in an unknown environment, and the path will adjust as the robot continues to explore.

Video please see the source blog

These are the path planning algorithms that most mobile robots need at present, and as one of the first service robots appearing in the consumer market, the path planning algorithm is more complicated.

Generally speaking, the sweeper needs so many planning ability: the Welt sweep, the I-sweep of the turn back and the power when the electricity is charged independently . These basic needs cannot be met by relying solely on the d* algorithms described previously.

Sweeping robots also require additional planning algorithms, such as I-cleaning for reentry, with many problems to be dealt with. How to sweep the machine the most effective cleaning without repeated cleaning? How to let the sweeper and people understand the concept of the room, door, corridor?

In view of these problems, academia has long been a specialized research topic called space coverage , but also put forward a lot of algorithms and theories. Among them, the more famous is Morse decompositions, sweeping machine through it to achieve the division of space, followed by cleaning.

In the 1970s, Carnegie Mellon University (CMU) relied entirely on ultrasound to do what we do now, and of course the cost is very expensive.

Moving path planning from point A to point B described earlier is also the basis for such a more advanced path planning. In fact, there is still a lot of work to be done from slam to the functions needed to sweep the robot. For the sweeping robot, we have its unique path planning function pre-built in the Slamware, convenient for manufacturers to integrate, do not need two times development.

This article turns from Lei Feng Net, if need to reprint please to Lei Feng Network official website to apply for authorization.

The secret of autonomous movement of robots: What is the relationship between slam and path planning? Three