KR100715609B1 - Topological mapping and navigation method of mobile robot in home environment - Google Patents

Abstract

The present invention relates to a topological map formation and driving method for a mobile robot in a home environment to efficiently map and autonomously travel.

The present invention provides a topological map formation and autonomous driving method of a mobile robot in a home environment.

The topological map formation and autonomous driving method,

Using an edge having a width that can be passed by the mobile robot but having a limited width, and a node which is surrounded by obstacles and which can move to another node through the edge; The topological map forming method recognizes an edge using information from sensors and forms a topological map around the edges;

In order to form the topological map,

Basically move along the wall to locate the edges, divide the node area, and form a map for the node area; After moving to another node through the edge, it is characterized by repeating the process of removing the wrong edges.

Using the present invention as described above, there is provided a topological map formation and driving method for the mobile robot to efficiently map and autonomously travel in a home environment.

Home Environment, Mobile Robot, Mapping, Autonomous Driving

Description

Topological Mapping and Navigation Method of Mobile Robot in Home Environment

1 shows a topological structure according to a first prior art

2 shows a topological model according to a second prior art

Figure 3 shows a topological map according to the third prior art

4 shows a general home environment (a) and a topological map (b) in the home environment.

Fig. 5 shows the definition (a) of the edge used in the present invention and the edge area (b) in the home environment.

Figure 6 shows the case of edge (a) and the case of not being edge (b) in the present invention.

Fig. 7 shows a general home environment (a) and a node area (b) in the home environment as applied in the present invention.

8 shows two types (a) and (b) in the case of recognizing edges in the present invention.

9 illustrates cases in which an edge is generated in the present invention.

Figure 10 shows a flow chart (a) of moving to a new node and a flow chart (b) of a new node in the present invention.

Figure 11 shows the overall flow chart of the movement of the mobile robot in the present invention.

Fig. 12 shows the indoor environment (a) in the first simulation and the topological structure (b) in this environment.

Figure 13 shows Node 1 (a), Node 2 (b), Node 3 (c), Node 4 (d), Node 5 (e), and Node 6 (f) in the local grid map of the first simulation of the present invention. Shows

Fig. 14 shows the indoor environment a in the second simulation of the present invention and the topological structure b in this environment.

Figure 15 shows Node 1 (a), Node 2 (b) and Node 3 (c) of the second simulation of the present invention.

Fig. 16 shows the indoor environment (a) and the topological structure (b) in this environment for the experiment on the edge case 1 in the present invention.

Figure 17 shows nodes 1 (a) and 2 (b) of the local grid map in the experimental environment of edge case 1 in the present invention.

Fig. 18 shows the indoor environment (a) and the topological structure (b) in this environment for the experiment on the edge case 2 in the present invention.

Figure 19 shows nodes 1 (a) and 2 (b) of the local grid map in the experimental environment of edge case 2 in the present invention.

Fig. 20 shows the indoor environment (a) and the local grid map (b) in this environment for the experiment on the edge case 4 in the present invention.

Fig. 21 shows the indoor environment (a) and the local grid map (b) in this environment for the experiment on the edge case 5 in the present invention.

FIG. 22 shows, in the present invention, an indoor environment (a) for experiments on edge case 6 and a local grid map (b) in this environment.

Figure 23 shows, in the present invention, an indoor environment (a) with two nodes and one dead-end edge, and a topological structure (b) in this environment.

Figure 24 shows nodes 1 (a) and 2 (b) of the local grid map in Figure 23, in the present invention.

The present invention relates to a topological map formation and driving method for a mobile robot in a home environment to efficiently map and autonomously travel.

With the recent development of mobile robot technology, mobile robots are no longer used only in industrial sites and are being distributed to various fields, and mobile robot technologies are applied to military purposes from robots that can be seen at home such as cleaning robots and educational robots.

However, the current autonomous driving technology is not only at the level of performing a preprogrammed function but also at the level of avoiding obstacles, and thus does not satisfy the autonomous driving at all.

This can be seen in Rumba (origin: USA) or trilobite (origin: Sweden), a low-cost cleaner robot that has been widely used in homes recently. These robots move along a preprogrammed path in a given area, cleaning up avoiding obstacles and do not form an accurate map.

In fact, the general household environment is different from each other and is quite complicated, so it is difficult to properly clean and move along a pre-defined path in such a home environment, and it is virtually impossible to clean between rooms. .

To solve this problem, E. Gonzalez, A. Suarez, C. Moreno, and F. Artigue, "Complementary regions: a surface filling algorithm," in Proc. IEEE International Conference on Robotics and Automation, 1996, pp. At 909-914, an attempt was made to construct a topological structure in the home environment, as shown in FIG.

This study constructed the topological structure as shown in (b) in the environment as shown in FIG. However, this configuration is divided according to how the robot moves. Therefore, the configuration is greatly changed depending on the obstacles or environment that the robot encounters. This topological structure does not mean much.

See also N. Tomatis, I. Nourbakhsh, R. Siegwart, `` Simultaneous localization and map building: a global topological model with local metric maps '', in Proc. IEEE International Conference on Intelligent Robots and Systems, 2001, pp. 421 In Fig. -426, as shown in Fig. 2, we studied how to construct a topological map and a local grid map in a home environment.

These studies are for constructing hybrid maps. In other words, it is important to obtain important feature points (such as corners) at each location and to organize the locations into topological maps. In this study, the corner points are found by the laser sensor, and the topological map varies according to whether the corner points are recognized.

And in P. Althaus and H. I. Christensen, "Automatic map acquisition for navigation in domestic environments," in Proc. IEEE International Conference on Robotics and Automation, 2003, pp. In 1551-1556, as shown in Fig. 3, a topological map as shown in Fig. 3 (b) is obtained in an environment as shown in Fig. 3 (a). This study has many rooms centered on the corridor and will connect them with a topological map. The fatal weakness of this study, however, is that each door must inform the user as a door.

The present invention, in view of the problems of the prior art as described above, in order to have a mobile robot has the ability to form a certain accurate map by themselves according to the changing home environment and to clean based on the map, driving from the map formation It provides an algorithm that can autonomously perform all processes up to.

To this end, the present invention proposes an algorithm that can form a map by crossing between rooms in a home environment and cleaning based on the map.

Home environments usually have grid maps, but by defining new nodes and edges, these environments can be broken down into topological maps and grid maps and have more advantages in routing, as well as cleaning critical sections or cleaning specific areas in the cleaning robot. Has the advantage.

More specifically, the present invention finds locations that must pass in order to pass to other areas in the home environment, constructs a topological structure around those locations, and points that must be passed by one edge. The topological structure is represented as a topological structure, and the obtained topological structure does not generate many nodes and edges, and the configured topological structure is not easily changed by obstacles.

As will be described in more detail below, the basic concept of the present invention may be referred to as a basic concept to divide a home environment into a large area such as a room. In this way, when divided into areas, it is advantageous to move the robot to a specific area. In other words, since the spatial concept is similar to that of ordinary people (divided into rooms), the commercial robot can be moved to a specific room.

In addition, in the present invention, the robot constructs a topological map by autonomously performing all processes. The robot moves on its own until the end of the entire mapping, resulting in a local grid map and a topological map of each area.

The general home environment is composed of a large living room and several rooms as shown in FIG. Each room and living room is divided into walls or fixed obstacles, and each room has a specific space, or door, that moves to another room or living room.

Taking advantage of these characteristics, if the home environment is divided into several areas, it can be divided as shown in FIG. The areas may be a room that ordinary people think of, or they may be divided by some fixed obstacle, like dividing the living room into two areas. Each area has a characteristic area that can be moved to another area, which can be a door or an area created by a fixed obstacle. If the entire home environment is divided into several areas around this area, Fig. 4 (a) is divided into four areas.

In addition, if such divided areas are called nodes and areas such as doors connecting the nodes are called edges, the topological structure may be configured as shown in FIG.

There are many reasons for introducing topological structures into the home environment. First, by defining nodes and edges, we know where to go when moving from one area to another, and we can use these nodes and edges to form a path for movement. And it has the advantage of having a separate map of the home environment.

<Node and Edge Definition in Home Environment>

Before we define nodes and edges, let's start with some assumptions. These assumptions are essential to map formation.

1. It is possible to cross between rooms in a home environment. There is no threshold and the visit is open.

2. Position recognition is possible.

3. The size of the home environment space to be mapped is fixed.

4. Obstacles are not moved during map formation, and obstacles can be moved after map formation. That is, there is no dynamic obstacle in map formation.

5. When forming a map, the robot is not blocked by obstacles and cannot move to another area.

The first assumption is that the robot must be physically mobile. In the past, the door had a threshold, making it difficult for the mobile robot to cross it. That is, mapping and driving were possible in only one room, but recently, most houses are removing the threshold, so if a visit is open, the robot can move to another room. This assumption is necessary because the robot must map the entire home environment.

The second assumption is that position recognition should be possible. This is because it is impossible to form an accurate map without location recognition. Location recognition can be done using the latest barcoded flooring materials or with StarLite from ETRI. That is, it is performed in an environment where such location recognition is possible. In the present invention, the location recognition using a barcode and based on the location recognition and driving and map formation algorithm is applied.

The third assumption is that the area to be mapped should not be infinitely large. When the topological mapping is composed by Depth First Search method, the topological map is completed after the map formation is completed. If the area to be mapped is infinitely large, it is impossible to complete the mapping process, and therefore a topological map cannot be obtained. Therefore, the area to be mapped should be limited.

The fourth hypothesis is necessary because the edges and nodes will be different if obstacle movement occurs during map formation.

Finally, the assumption that the robot is not blocked by obstacles and cannot move to another area is similar to the first assumption, to prevent the robot from being blocked by obstacles when the robot moves to another area. If the robot is unable to move because of an obstacle, the robot cannot finish mapping all areas.

In general, nodes in a topological map are characteristic and meaningful points, and edges are known to connect nodes. However, in the present invention, since the edge is more meaningful than the node, the edge is defined first, and then the node is defined.

The definition of edge is as follows (see Figure 5 (a)).

Definition 1. (edge)

The location that connects one area with another area, the width of which the robot can pass, but the location with a limited width is called the edge. The edge satisfying the following is an edge.

An empty area between static obstacles that the robot can pass through.

Width W between static obstacles.

(3/2) R <W <L (R: radius of robot, L: width of door)

· Connections between nodes

Usually the door width is fixed in each country in the world. In most countries, from 0.7m to 1.5m, 1.5m was chosen as the value of L in the present invention. The value of L is not greater than the door width here. But that's not just the door. Areas corresponding to the above definition are all edges (see Fig. 5 (b)). Edge has the following information.

· Both connected node numbers (index of Istart node and end node)

Edge start and end positions

Relative position between connected nodes (relative position between start and end nodes)

Edge direction

The position of the edge is the position in the middle of the width of the edge. The direction (edge direction) is a direction in which another node is in the direction that the edge is facing. In other words, it is the direction perpendicular to the part recognized as the edge and the direction from the start position to the end position of the edge.

When defining edges like this, edges have several characteristics.

Temper 1. (Edge)

• Both ends of the edge have different areas (nodes).

If the nodes connected at both ends have the same indices and their positions are the same edges.

There can be multiple edges connecting one area.

Both ends of the edge must be connected to different areas because both ends of the edges defined above cannot be the same area. And it means that both ends must lead to some area. This is because both ends of an area cannot be the same in space. That is, the case of Fig. 6 (a) becomes an edge and the case of Fig. 6 (b) can never come out. In addition, although there may exist several areas connected from one space to another space, several same areas may not exist at the same time until the location of the connected area is the same. If the nodes connected at both ends have the same indices and their positions are the same, they are bound to be the same edge.

In a typical home environment, the space is divided by walls. These spaces are surrounded by walls and can be moved to other rooms by doors. Nodes are spaces surrounded by walls or fixed obstacles (see Fig. 7 (a)). The exact definition of this node is as follows:

Definition 2. (Node)

Nodes are spaces that can be moved to other nodes through the edges in areas surrounded by walls or fixed obstacles.

The node meets the following conditions.

Area surrounded by static obstacles or walls

Connected to other nodes by edges

The node has the following information:

Node number (node index)

Linked edge information

Map information (local grid map)
When defining a node as above, the node has the following properties:

delete

Temperance 2. (Node)

Each area has independent spaces

The nature of these nodes is that the spaces cannot overlap each other. Because each node is surrounded by a wall and divided by that wall.

Since a node can be thought of as an area like a room, a home environment can be divided into several nodes. Since each node has its own grid map as shown in Fig. 7 (b), it is possible to prevent errors accumulated at each node from accumulating continuously. If the map information for the whole home environment is in one grid map, the errors are insulated for the whole area. However, if you have a grid map of the area corresponding to the nodes using topological structure, the errors in each node are independent of each other.

The robot basically has to do the initial mapping while moving along the wall. This is the pre-mapping process described later, in which the robot can recognize edges connected to nodes. In this case, the edge recognition is divided into two cases as shown in FIG. Type A is where the edge is on the right side of the robot running along the wall and type B is when it is in front of the robot.

In the case of type A, when moving along a wall and rotating to the right, the distance between objects is within the distance corresponding to the definition of a given edge, and there is an edge when no object is recognized between them. (This can be checked with a sensor. In our experiment, we used 7 PSD sensors, and the distance between the sensor in the -30 degree direction and the position perceived as the sensor at +90 degree belongs to the distance corresponding to the given edge definition. The sensor in between was recognized that there is an edge when no object is detected. You can use the distance sensor to know the width of the object even if it is not a PSD sensor.)

 Type B, on the other hand, occurs in all directions while the robot is moving, and is the edge that appears when the robot is moving along the wall or moving straight ahead. (In the experiment, the distance between the sensor at -60 and -90 degrees and the position perceived as the sensor at +60 and +90 degrees is the distance corresponding to the edge definition, and the sensor in between is not recognized. You will recognize that there is an edge.)

Depending on the type of edge, the movement of the robot is slightly different. The movement will be explained in detail later.

<Cases where edges are created>

As mentioned earlier, constructing a topological structure in a home environment by defining new nodes and edges can have many advantages. However, in many parts that do not need to be defined as edges, many parts are often defined as edges (see Fig. 9).

FIG. 9 (a) shows a case where two edges passing to another node exist. For example, there are two doors connecting a room to a room. This corresponds to the nature of the edge, in which case the robot creates two edges.

Fig. 9 (b) is a case where the edge is originally one but the obstacle is located between the edges, and the robot thinks that there are two edges. In this case, the robot determines that there are two edges, but this is not a problem at all. This is because the robot can move to the other edge using either of them.

9 (c) is a case where an obstacle is placed between the edges as shown in FIG. 9 (b), but the robot recognizes that there is only one edge. In this case, there is no change except that the position of the existing edge is slightly changed. Therefore, the robot can move to the recognized edge.

In FIG. 9 (d), there is one edge but the robot does not recognize the edge because an obstacle is placed between the edges. The difference from FIG. 9 (b) is that the width is narrower than that of the existing edge in FIG. 9 (d). In this case, the edge is so narrow that it is judged that there is no edge, and the fact that it is not recognized as an edge because of the narrow width is not only able to pass by the robot, but there is also a risk of collision if it tries to pass. In other words, it is very dangerous for a robot to move to another node between them, contrary to what was initially assumed (assuming that the robot should not move to another area).

9E illustrates a case where no edge exists. But the robot thinks that there are two edges. This is because the robot recognizes two edges in the same position. However, these two edges are not really edges. Both edges have identical nodes connected at both ends. This is contrary to the nature of the edge. This is because the nodes connected at both ends must be different nodes. When the robot moves through an unexplored edge to move to a new node, it knows that these edges are on the same edge and on the same node.

In this case, two edges are removed. In other words, the robot initially recognizes that there are two edges, but eventually the edges are removed if they move to map other nodes. This removal method is described in more detail below.

Fig. 9 (f) is also a case where it is not an edge. However, the robot initially perceives an edge. However, we can see that we try to move to another node through this edge, but no other node is connected to this edge. We will define this edge as a dead-end edge. Dead-end edges must be removed from the topological map.

Cyclic cases are shown in Figs. 9 (a) and 9 (b). This case can be solved by checking whether a new node matches a previously matched node. In the nature of nodes, nodes have spaces that are independent of each other. In other words, they cannot overlap each other spatially. Robots have a topological structure but also have relative positions between nodes. Therefore, when the robot goes to a new node, the relative position can be used to determine if this node is a previously registered node. If the same node exists before, the current edge is found and merged with the previous node without adding it as a new node.

<Autonomous Driving for Map Construction>

We will describe the overall process of autonomous driving by using the definition, creation and processing of the edges described above. The overall process of autonomous driving is to create the topological structure in the home environment through the recognition of edges, the elimination of unnecessary edges, and the integration process for the same edges and nodes.

After constructing the topological structure, when the robot moves from the position of one node to the other node, it is much more efficient in the path formation and the errors in each node are independent of each other, thus preventing the error in the map formation. And even in map formation, it is much more efficient in map formation because it finishes the mapping of one node and moves it to another node. The autonomous driving method for constructing the topological structure is as follows.

1. Node Mapping

(1) pre-mapping

In this process, the robot basically moves along the wall. When the robot moves along the wall, the robot recognizes the edge. The method of recognition has been described above. If the edge continues to move along the wall from where it was recognized, it will fall to another node. This results in basic wall-following and no topological structure.

When an edge is detected, it must move in a direction perpendicular to the direction of the edge. This rule applies to both type A and type B. In the case of type A, the robot is located on the right side of the robot's position. Therefore, the robot can move by rotating in the direction perpendicular to the direction of the edge. If you can no longer move by an obstacle or wall in front of you, or you can move along the wall again, move along the wall again. However, in the case of Type B, if you move in the same way as Type A, you will be transferred to another node. In the case of Type B, you must first stop, rotate in a direction perpendicular to the edge direction, and then move along the wall when you can move along the wall. If you move along the wall, the robot will move along the edge to another node, but if you do this, the robot will not move to another node until the pre-mapping is complete. After the pre-mapping, the robot knows the number of candidate edges, the position of the starting position of the edge, and the overall size of the node.

delete

(2) Full-Mapping

Full-mapping is the process of mapping the entire area of a node. In the present invention, a Zig-Zag algorithm is used. However, the zigzag algorithm is a bit inefficient to cover the whole area. This is because there may be obstacles in the middle of the node. In addition to the zig zag algorithm, other efficient coverage algorithms can be used during the full-mapping process. However, this process is not an important subject in the present invention and will not be discussed further. However, in order to apply the zigzag algorithm more efficiently, it is good to find the most corner point in the area and apply the zigzag algorithm from that point. Therefore, in the pre-mapping process, each corner point is saved, and before the pre-mapping is finished and before the full-mapping, the robot finds the most suitable position among the corner points, and moves the robot to that position, and then performs the full-mapping from there. To start.

2. Autonomous Driving for Topological Mapping

(1) move to new node

After full-mapping at a node, the robot checks to see if there are unexplored edges at the current node. If there is an unexplored edge, the robot will move to an unexplored edge. At this time, the moving point is the starting point of the edge which has not been investigated yet, and the robot moves to the A * algorithm (algorithm for the governor as an algorithm that shows how to form the shortest distance from any starting point to the end point).

When the robot reaches the start of the edge, it moves along the edge until a new node appears. When moving along the edge, it should be checked whether it is a case of Fig. 9 (e), Fig. 9 (f) or a cyclical case. First, as shown in the flowchart of Fig. 10 (a), it continuously checks whether a new node is coming out or can no longer move along the edge. If the robot can no longer move along the edge, the edge is shown in Fig. 9 (f). A dead-end edge with no other node connected at one end is removed from the topological map.

When a new node appears, it is checked whether or not it is the case of FIG. When a new node appears after moving along an edge, the position becomes the end point of the edge. If the position of the end point coincides with the position of the start point of another edge of the immediately preceding node, this case is shown in Fig. 9 (e). As is the case, it should be a Fusion edge and both edges should be removed from the Candidate edge. If it is determined that the case is not the case of Fig. 9 (e), it is necessary to confirm whether the new node is a node previously present in the topological map. This case appears in the cases of Figs. 9 (a) and 9 (b).

Due to the characteristics of the edge and the node defined in the present invention, there may be several edges connected between the nodes. If you do not verify that the node is a previous node, it will continue to be added as a new node. In this case, the node grows indefinitely, so it is necessary to check whether the node came out before. Although it has a topological structure in the home environment, each node has a relative position and these positions do not change, so each node cannot overlap. Therefore, the relative position allows the robot to determine if it is a node that came out before.

If it is determined that it is not a previously released node, the robot adds a new node to the topological map, the robot initializes its current position to (0, 0), forms a new local grid map, and starts the pre-mapping process. You will start again.

(2) move to final node

In the present invention, the Depth First Search algorithm is used as a method of constructing a graphical structure to construct a topological map. Therefore, if there is no unexplored edge yet, the robot moves to the final node (commonly called the parent node). In this process, the robot finds the edge moving to the final node and moves to the starting point of the edge. The movement algorithm uses the algorithm A * used in the process of moving to a new node.

After moving to the last node, the robot checks again for unexplored edges, and if so, it goes back to the new node and finds a new node. If this node has no edges yet to be investigated, it will move back to the final node. If there are no more final nodes while moving to the last node, the mapping process ends. When the node that started the mapping is called node 1, the robot searches for all nodes and then returns to node 1, and when the edge is no longer investigated, the whole process is finished. The overall autonomous driving flowchart is shown in FIG.

Advantages of Topological Maps in the Home Environment

When the map is formed in the home environment using the topological structure, it has the following advantages.

When moving from one node location to another node, there is no need to search the entire map to form a route, find the node index connected from the current node to the node that you want to go to, and then go to the edge where each node has the node index connected. You only need to form a path. This is very efficient computationally. Searching the entire map in route formation can take as little as a few seconds to a few minutes. However, connected nodes index search takes very little time and takes much less than searching the entire map to form a route from each node to the edge.

2. And, even when forming map, it is efficient because it finishes mapping of one node and moves to another node to form the map. There are usually walls and many fixed obstacles in a typical home environment, which is very complicated and inefficient when one coverage algorithm or an Exploring algorithm is applied. On the other hand, if you finish one node and move to another to map it, it is much simpler to apply the same algorithm.

3. Next, the errors in each node are independent of each other. Each time you move to each node, you use a map corresponding to each node, so errors in each node have a limited effect on each node. This prevents the error from increasing in map formation.

4. Finally, the topological structure constructed commercially is similar to the spatial concept that ordinary people think. Similar to the idea that the common people are divided into a room and a living room, they get a topological map, so the robot can be moved to a room desired by the user, thereby increasing the area of commercial use.

Simulation Results

In order to show the autonomous driving method for the topological structure in the home environment, the simulation was first performed in the environment as shown in FIG. In order to make the simulation environment similar to the real environment, the simulation was performed by reducing the radius of the robot to 210mm, the size of the home environment to 8mx12m and the thickness of the wall to 100mm, as shown in Fig. 12 (a).

In the simulation, there are fixed obstacles in the middle, each room is surrounded by walls or fixed obstacles, and there is no physical problem to pass the visit. And it is assumed that areas such as the multipurpose room and the restroom cannot go. The environment as shown in Fig. 12A is divided into six nodes.

In FIG. 12 (a), the case 1 may be referred to as the case of FIG. 9 (c). It is the case that the edge is existing but the position of the edge has changed because of the flower pot. In this case, there is an edge between the wall and the fixed obstacle. Case 2 corresponds to Fig. 9 (f). Initially, you think an edge exists in Case 2, and when you try to move to another node through that edge, you know that no other node is connected to this edge. This does not fit into the definition of an edge, which is called a dead-end edge and is removed from the topological map. Case 3 is the case shown in Fig. 9E. In this case, the robot initially thinks that there are two edges, and then knows that the same edges exist when moving to another node. These two edges are found to be fusion edges, not real edges, and are removed. The overall topological structure is as shown in Fig. 12 (b). Local grid map information of each node resulting from the simulation is shown in FIG.

The number of edges connected to each node is 4 nodes 1, 2 nodes 5, and 1 node 2, 3, 4, 6. The results show that Case 2 and Case 3 were not treated as edges. To construct the topological structure, we used the Depth First Search method and the node process that the robot moved is as follows.

Node 1 → Node 2 → Node 1 → Node 3 → Node 1 → Node 4 → Node 1 → Node

5 → Node 6 → Node 5 → Node 1

<Autonomous Driving Experiments and Results for Constructing Topological Maps>

This experiment was carried out in a location-aware environment (experimental in a floor with a barcode). As can be assumed, the experiment can be performed in a location-aware environment. If the location is not properly recognized, you will not get an accurate map. The experiment can be divided into two types. The first is to check whether the nodes can be formed and the topological map can be obtained. The second is to test whether the edges can be handled as described above.

1. Topological Mapping

First, we experimented in an environment consisting of three nodes. The experimental environment is as shown in Fig. 14 (a), and the topological structure in this environment is as shown in Fig. 14 (b). In this environment, the robot first maps the first node, and once the mapping is complete, the robot knows that there is an unexplored edge, so it moves through the edge to the new node. When a new node is recognized, the second node is registered and mapped.

After the mapping process of the second node is over, the second node knows that there is one edge that has not yet been investigated and one edge that has been explored, and then moves to the new node through the edge that has not yet been investigated. Then map the third node. After mapping to the third node, we know that there are no more edges that have not yet been investigated, and as we move to the parent node, we continue to search for edges that have not yet been investigated. The mapping process ends because it no longer exists. The local grid map of each node obtained in this mapping process is shown in FIG.

(1) Experiment with edge case

9 illustrates various edge cases.

We will experimentally verify the treatment method for each case described above.

<Case 1>
As illustrated in FIG. 9 (a), a plurality of edges connecting the room to the room may exist in FIG. 16 (a). In Fig. 16 (a), there are two regions, such as doors, which can be crossed between the two regions, and the topological structure in this case is as shown in Fig. 16 (b). In this case, it becomes a cyclic loop in the topological structure. If you don't recognize them as being the same node, you end up in an infinite cycle. Again, after mapping the first node, we can see that there are two edges that have not been investigated yet. The robot goes to node 2 through one of the edges that has not yet been investigated. After mapping node 2, I think there is one explored edge and one unexplored edge. After mapping node 2 is finished, it goes back to node 1 through the unexplored edge. However, the robot recognizes that it is going to a new node instead of node 1. In other words, if you don't know that this node is the first node that came in, it will be added as node 3 and it will continue into infinite cycle. However, as explained earlier, each node has the property of having independent spaces. Each edge has a relative position of the connected nodes. Therefore, when a robot comes to node 1, it knows that it is not a new node, but is node 1, which it had previously mapped, and does not add a new node. It updates the information by searching for which edge is the same at node 1. It then moves to the next edge or node according to the Depth First Search method. After the experiment, the local grid map of each node is shown in FIG. 17.

delete

<Case 2>
Fig. 18A shows the case where there are two edges as shown in Fig.17. Originally, as shown in Fig. 18 (a), there was only one edge, but there was an obstacle in between to change the case to two edges. As mentioned earlier, even if the edge is increased from 1 to 2, there is no problem moving to another node regardless of which edge is moved. The topological structure is shown in Fig. 18 (b), which is the same as the case 1 described above. The map formation method is the same as the case 1 described above. The local grid map of each node obtained as a result of the experiment is shown in FIG. If you look at the local grid map, you can see that two edges are caused by obstacles.

delete

<Case 3>

As in Case 2, there is an obstacle between the edges, but the robot recognizes that there is only one edge. In this case, there is no change except that the position of the existing edge is slightly changed. Therefore, the robot moves to the recognized edge.

<Case 4>
Case 4 is a case where the robot cannot be moved to another node because the robot is blocked by an obstacle as described in FIG. 20 (a). In Fig. 20 (b), although there are two nodes and there is one passage connecting between the nodes, there is an obstacle between the passages so that the robot cannot cross between the nodes. In other words, the edge existed, but it was blocked by an obstacle, and the edge disappeared. In this case, the area that the robot needs to map remains, but the robot no longer maps and thinks there is only one node. As shown in Fig. 19 (b), there is a small space between the obstacle and the wall, but it is so narrow that the robot cannot pass through the space. Since the above assumes that the robot should not be blocked by obstacles and cannot move to other areas, such cases should be excluded when forming the map.

delete

<Case 5 (fusion edge)>
Cases 5 and 6 are very unusual among the cases where edges can occur. First, in case 5, the edge does not exist as shown in Fig. 21 (a), but the edge is considered to exist due to an obstacle in the middle. As the robot moves along the wall, it thinks an edge exists between the wall and the obstacle, or between the obstacle and the obstacle, but in reality it is not an edge. We want to define this edge as a fusion edge. Fusion edges are always present in pairs. In the nature of an edge, an edge must always be connected to another node. Fusion edges have the same nodes connected at both ends. The robot tries to move to the new node through this edge, but the new node does not come out and recognizes that it is its own node. Fusion edges exist in pairs, with their start and end points reversed. When recognized as a fusion edge, both edges are removed from the topological structure. Therefore, in the case of Fig. 21 (a), only one node exists and the result is as shown in Fig. 21 (b).

delete

<Case 6 (dead end edge)>

Case 6 is the same as in Fig. 22A. In this case, we can say that it is a dead-end edge. As shown in Fig. 22 (a), there is one node and there is one dead edge. This edge is an edge that is only connected to this node, not to other nodes. As explained above, if a node is not connected at both ends, it is not an edge. The robot will end up mapping the node and move to the dead-end edge and move over to the new node in order to move to another node. The robot knows that this edge is a blocked dead-end edge, Should be removed. As a result, Fig. 22 (a) shows that there is no edge connected with one node. The local grid map obtained as a result of the experiment is shown in Fig. 22 (b).

<Two Nodes and One Dead-End Edge>

We experimented with Case 6 earlier, but for a clearer result, let's talk more about how the robot moves and handles Dead-End edges when there are two nodes and one dead-end edge. 23 (a) shows an edge connecting two nodes and two nodes, and a dead end edge connected to the first node. The topological structure in this case is constructed as shown in Fig. 23 (b). The robot recognizes that there are two edges after the first node mapping, and passes to the second node through the first edge. After mapping the second node, we know that there are no edges that have not been investigated yet, and we move on to the first node, the final node. We tried to move from the first node to the second edge, an edge that hasn't been investigated yet, and move on to the new node, but the second edge is not connected to another node as a dead-end edge. This edge is removed from the topological structure. The local grid map obtained after the experiment is shown in FIG.

After mapping, the robot can move efficiently with the topological map. It is also possible to know which location is located at which node, so that the topological map can be used even when moving.

In the present invention, by defining new nodes and edges, an algorithm that can construct a suitable topological structure in a home environment is proposed. It defines the proper nodes and edges by taking advantage of the fact that a normal home environment is surrounded by walls or fixed obstacles and must pass through an area such as a door when moving from one area to another. By this definition, non-edge regions can be defined as edges. By clearly distinguishing and defining the characteristics and properties of edges, it is possible to construct a topological structure suitable for home environment.

When autonomous driving using topological structure finishes the mapping of one area and moves on to another area, it is easier to apply the coverage algorithm or the mapping algorithm, and the errors in each node are independent of each other. It can prevent the increase. In addition, it has much advantage in terms of calculation speed because it has information about the edges, that is, edges, which must pass when the robot forms a path after the mapping process is completed.

In addition, the topological structure constructed by the present invention is similar to the spatial concept that ordinary people think, so that the robot can be moved to a room desired by the user, thereby increasing the area for commercial use.

Claims (8)

In Topological Map Formation and Autonomous Driving Method of Mobile Robot in Home Environment, The topological map formation and autonomous driving method, Using an edge having a width that can be passed by the mobile robot but having a limited width, and a node which is surrounded by obstacles and which can move to another node through the edge; The topological map forming method recognizes an edge using information from sensors and forms a topological map around the edges; In order to form the topological map, Basically move along the wall to locate the edges, divide the node area, and form a map for the node area; A method of topological map formation and autonomous driving of a mobile robot in a home environment, characterized by repeating a process of removing an invalid edge after moving to another node through an edge. The method of claim 1, wherein the moving to another node through the edge uses a Depth First Search method. The number and position of edges connected to an area defined as a node by moving in a direction parallel to the wall when the robot locates the edges while moving along the wall in autonomous driving to form a map. The topological map formation and autonomous driving method of the mobile robot in a home environment, characterized in that, to determine the topological map, the number and location of the identified edges are used. The method of claim 2, wherein the movement to another node through the edge is performed. Topological map formation and autonomous driving method of mobile robot in the home environment, characterized by finding the edge to be moved among the connected edges, moving to the starting point of the edge, and moving to the next node by moving to the edge direction. The method according to claim 1, wherein, in the case of mapping, when two or more edges are connected or when two or more edges are recognized as obstacles lying between each other, whether or not to add a node is determined by a relative position between each node. Topological map formation and autonomous driving method of mobile robot in home environment.  According to claim 1, In order to remove the wrong edge, the edge of the mobile robot in the home environment characterized in that the edge does not exist, but the edge is considered to have an edge between the obstacle and the obstacle. Topological Mapping and Autonomous Driving.  2. The method of claim 1, wherein in the case of a dead-end edge where no other node is connected at one end to remove the wrong edge, the topological map formation of the mobile robot in the home environment is removed. Autonomous driving method.  7. The method of claim 6, wherein the fusion edge removal assumes that the end point of the edge when a new node appears after moving along the edge is removed if it coincides with the position of the start point of the other edge of the immediately preceding node. Topological Mapping and Autonomous Driving Method of Mobile Robot in the Environment.

Images