JP3218865B2 - Route search device incorporating time axis into search space - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a route search device for a mobile robot incorporating a time axis into a search space. Mobile robots that are currently in practical use, such as parts transport vehicles in factories, automatically operate on a fixed schedule on a fixed course guided by a line drawn on the floor in a well-managed environment. There are two types, one that moves by remote control in a bad environment where people cannot go, and one that moves autonomously. It relates to a search device.

[0002]

2. Description of the Related Art FIG. 6 is a diagram showing an example of a configuration of a conventional mobile robot. A bold frame indicates an entire configuration, and a dotted frame indicates an example of a configuration of a route generation unit.

A mobile robot is equipped with cameras 13a and 13b such as an ultrasonic sensor and a touch sensor for observing environmental conditions. When performing remote control, a human moves the mobile robot based on information from the sensors 13a and 13b.

In order to perform autonomous movement instead of the remote control described above, a route generation unit 14 is provided based on a map of an environment in which a mobile robot moves about in advance.
In, a route plan is created, and a moving operation based on the route plan is executed. Specifically, based on the route information determined in the route generation unit 14, the central control unit 10
The drive control unit 11 is activated, and the drive device 12 moves along the determined route.

[0005] The method of creating the route plan is performed as follows. In other words, the cost of an environment map ｛grid or cubic grid unit that describes the existence of obstacles such as walls and immovable objects and the time required Based on the described cost map｝,
Calculate the cost, such as the total time required for several routes,
Various routes that satisfy constraints such as reachability are found, and a route with the minimum cost is determined from the routes. When actually moving around, there are obstacles that are not on the map and moving objects such as humans and other mobile robots, and collision avoidance of those that are not on the above route plan is performed by the central control unit.
The operation at 10 is performed as an exception process, and a collision avoiding operation, for example, a detour operation, which is determined separately from the moving operation along the route plan, is executed. If the detour operation is not possible, the process returns to the movement start point, determines another route, and moves.

As described above, conventionally, when planning a route, events in a map of an environment in which a mobile robot moves around are always time-invariant and the information obtained is always accurate without any missing information. It is a condition.

Therefore, knowledge of time-varying events, such as doors at entrances that open and close at regular times and moving objects that move at a constant speed, is used in the above route planning, specifically, based on a cost map. There is a problem that when finding a candidate and determining the route having the minimum cost among the routes, the route cannot be taken into the route plan. Therefore, when such time-varying events exist, there has been a problem that inconveniences such as erroneous determination of reachability and inability to reduce the cost to the true minimum occur.

Further, as described above, since the collision avoidance action is performed separately from the movement on the set route (ie, exception processing), there is a problem that the adjustment operation between the two kinds of actions becomes complicated. there were. For example, considering the return of the route after the collision avoidance operation, there is a problem that it is necessary to recognize the current position and the direction of the self-moving robot by new sensing and to set a new route for returning to the target route. .

[0009] Here, with regard to the correspondence to a moving object for which the motion can be predicted sufficiently accurately, if they can be incorporated into the path plan, the movement can be more efficient than the movement operation by collision avoidance as a mere exceptional processing. It should be.

Therefore, in order to perform a route search including events that change with time, a time axis is added to the two-dimensional or three-dimensional space axis to extend the route search space. The basic idea is as follows, and disclosed in detail by Japanese Patent Application No. 06-017702, "A route search method and apparatus incorporating a time axis into a search space", which was filed by the present applicant. I have.

First, consider the case where all the events in the environment have complete knowledge and the prediction is accurate. First, consider the effect of a case where the mobile robot has complete knowledge of all the events in the environment in which the mobile robot moves around and accurately predicts the behavior of the moving object. Comparing a search space that adds a time coordinate axis to the coordinate axes of a search space that includes events that do not change over time and a search space that includes only spatial axes that include only events that do not change over time, Can be treated the same, except for the following:

That is, since the time cannot be reversed and the moving speed of the moving object has a maximum limit, in the search space including the time axis, the direction of the route setting is limited. The target point is determined by a combination of the destination point and the time limit.

[0013] By performing the route planning in this way, the collision avoidance action with the moving object can be processed in a unified manner with the route planning in the obstacle that does not change with time. In this method, collision avoidance can be incorporated at the time of route planning, so that an efficient movement can be achieved, and there is no need to reset the route plan after collision avoidance.

However, in reality, accurate prediction of the future is impossible. Thus, in the invention of the prior application, the uncertainty of the prediction is treated as an existence probability distribution. The cost used for the route planning is calculated by integrating the existence probability of the event and the influence of the risk or the like.

[0015] Regarding the prediction of the future, the most difficult problem is
It is the interaction between events that occur in the environment. Performing the prediction of the interaction requires an enormous amount of calculation compared to performing the single prediction, and therefore, the interactions other than those having a significant influence are ignored. Further, the interaction between the mobile robot and events that occur in the environment causes the problem that the contents of the search space change during the route search.

Therefore, the route search method is based on the premise that the contents of the search space are fixed. Therefore, when the contents of the search space change, it is necessary to execute the route search. Can not. In the environment assumed here, the direction in which the interaction works may be cooperative with the mobile robot, such as avoiding collision.
That is, if the interaction between the mobile robot and the event that occurs in the environment is not taken into account, the efficiency is low, but it should not be dangerous. In view of this point, in the route searching method and apparatus according to the invention of the prior application, the route searching is performed without considering the interaction between the mobile robot and an event occurring in the environment.

FIGS. 7 to 18 are diagrams showing examples of handling of events in a search space incorporating a time axis. FIG. 7 to FIG. Path search method and apparatus incorporating axes in search space ", showing an example of an expression form of each event in a search space including a time axis. FIG. 7 shows a case of a constant event 8 shows an example of expression in the case of a periodic change event, FIG. 9 shows an example of expression in the case of a traveling means of regular operation, and FIG. 10 shows a constant speed movement available at any time. FIG. 11 shows an expression example in the case of means, and FIG. 11 shows an expression example in the case where there are many constant-speed moving means available at any time.
Shows an example of expression of a constant-speed moving means called and used at an arbitrary time, FIG. 13 shows an example of expression of a moving object that can be accurately predicted (moving object of constant velocity motion), and FIG. 15 shows an example of expression of a predictable moving object (constant speed motion), FIG. 15 shows an example of expression when a stochastically predictable moving object branches, and FIG. FIG. 17 shows an example of expression in the case of replacing a branch of an object with a temporally fixed probability distribution, FIG. 17 shows an example of expression in a case where a stochastically predictable moving object is found by sensing, and FIG. An expression example is shown in a case where the traveling path of the robot is sandwiched between moving objects that can be stochastically predicted.

The configuration and operation of a route search method and apparatus in which the time axis is also incorporated in the search space will be described with reference to FIGS. Here, for convenience of explanation, the spatial axes are x,
Alternatively, x, y, and the time axis are represented by t.

Constant event: The constant event mentioned here is the presence of an object (obstacle) whose position does not change with time. In this example, if such an object is captured in a three-dimensional space including the time axis shown in FIG. 7, it can be treated as a wall or a pillar as shown in the figure,
A route search can be performed. Accordingly, in the illustrated search space, the mobile robot can move between the slits and move on the route with the minimum cost.

Periodic change event: A periodic change event is an event that is known to occur exactly at a fixed time. Examples are entrances that open and close on a regular basis every day, and intersections with traffic lights. FIG. 8 shows an example of an entrance that opens and closes on a regular basis. If such an entrance is regarded as a three-dimensional space including the time axis t, it can be treated as a window and can be searched for. Can be. Therefore, in the illustrated search space, the mobile robot can move on the route with the minimum cost through the window.

Moving means for scheduled operation: This is an example of a moving means that is regularly operated according to a timetable (timetable), such as a train. FIG. 9 shows an example of expressing such a moving means in a three-dimensional space including a time axis t. Thus, such a moving means, when viewed as the three-dimensional space, as shown in the tunnel, ie, after waiting for a predetermined time after riding on the moving means, after moving at a predetermined speed, A tunnel formed when the vehicle is stopped for a predetermined time can be handled in a simulated manner, and the mobile robot can move on a route with the minimum cost through the tunnel.

Constant-speed moving means available at any time: This moving means is a moving means that moves at a constant speed available at any time, such as a belt conveyor or a moving sidewalk.
If this moving means is regarded as a three-dimensional space including the time axis t, as shown in FIG. 10, it can be handled as a corridor or a tunnel with a restricted moving direction, and can be handled as a mobile robot. Can be moved using the moving means, and the locus of the moving path at that time is as shown in FIG. 10, for example.

[0023] The above-mentioned means of transportation can be approximately regarded as a means of transportation for scheduled operation that is continuously operated.
When viewed as a three-dimensional space including the time axis t, FIG.
As shown in FIG. 1 (however, in FIG. 11, it is shown in two dimensions for convenience), it can be simulated that many tunnels are in close contact.

Means of transportation to call and use at any time:
This moving means is a moving means called and used at an arbitrary time, such as an elevator. In this case, since the time from when the call is made available until it becomes available is uncertain, a probabilistic treatment is required. Since the use after calling is at a constant speed, it can be used at an arbitrary time as shown in FIG. 12, and can be handled in the same way as a moving means at a constant speed. Figure 1 shows the actual trajectory of the mobile robot.
It becomes like the solid line of 2. That is, the part using the moving means is divided into a part that calls the moving means and waits, and a part that moves at a constant speed by using the called moving means.

From the viewpoint of route planning, once the decision is made to use the moving means, there is no need to consider the details of the use assumption, and the distance that can be moved using the moving means, Only the required time (corresponding to cost) at that time becomes a problem. Therefore, as shown in FIG. 12, by using the apparent speed, these moving processes can be simulated in the same manner as the moving means operating at a constant speed available at any time. it can. However, the waiting time part is
Since it is uncertain, try to find the average waiting time,
It is necessary to stochastically handle the apparent moving speed indicated by the dotted line.

Accurately Predictable Moving Object: This moving object is moving according to a timetable or according to strict movement rules, and when it is detected by sensing, its subsequent movement is accurately determined. It is a moving object (obstacle) that can be predicted. Taking such a moving object as a three-dimensional space including the time axis t can be simulated as a columnar obstacle as shown in FIG.

Stochastically predictable moving object: Although such a moving object cannot be predicted accurately, a database 1 shown in FIG. The moving object can be stochastically predicted by the simulation in the simulation unit 2 based on the information. (See FIGS. 14 to 18) First, an example of a moving object that moves at a constant speed will be described.
When this moving object (obstacle) moving at a constant speed is captured in a three-dimensional space including the time axis t, FIGS.
Can be simulated as a cone or an obstacle composed of a combination thereof. 14 to 1 above
In FIG. 8, the cone is drawn as a clear cone of the boundary, but actually, as a clear cone of the boundary as described above, depending on the prediction accuracy of the moving object and the risk of contact with the moving object. It is necessary to selectively use the method of simulating and the method of simulating as a cone having an unclear distribution of the boundary line.

FIG. 14 shows an example of predicting a moving object (obstacle) moving at a constant speed. The simulated cone is due to poor accuracy of the sensed speed and uncertainty.

FIG. 15 shows that a moving object moving at a constant speed is
This is an example of prediction when there is a possibility of branching at a certain point. In an actual application field, the robot does not move freely in a plane, and when there is an obstacle ahead of the moving object (for example, a person or a train) in the traveling direction or when the vehicle reaches an intersection in a corridor, A branch of the prediction of the traveling direction occurs. The prediction of the movement after the occurrence of the branch is based on the total probability of the probabilistic movement prediction in each branch direction and the probabilistic prediction of the branch direction. ) Is calculated.

FIG. 16 shows an example of prediction in a case where there is a large number of possible branches of a moving object moving at a constant speed. Simulating all the possibilities of all branches is not realistic because the amount of calculation becomes enormous and the value of the probability of each branch decreases. If the value falls below the set threshold, FIG.
As shown in (1), it is necessary to replace the distribution with a temporally fixed probability distribution (specifically, a fixed probability value at which the mobile robot can move with respect to the above-described lattice and cubic lattice) must be set.

FIG. 17 shows a case in which when a branch of a moving object moving at a constant speed is predicted, the predicted moving object is found by sensing at a certain point in time and its position is determined. This is an example. In this case, when the position is determined, the subsequent prediction of the moving object whose position has been predicted is stopped, and the prediction simulation of the detected moving object, that is, specific Starts setting the value of the probability that the mobile robot can move with respect to the aforementioned lattice and cubic lattice.

FIG. 18 shows an environment in which the moving direction is restricted, such as a road or a corridor, in which the moving path of the mobile robot is sandwiched between moving objects (obstacles) whose moving position can be predicted stochastically. Is shown. In this case, it is possible to perform a route search for the mobile robot by simulating the obstacle as a tunnel that narrows as it goes ahead.

As described above, the route search method and apparatus using the time axis in the search space describe a map or information of a moving object, and are two-dimensional or used for the route search of the mobile robot. A time axis t is added to the three-dimensional search space axis, and a time-varying event is replaced with an event on the space axis and processed. An object whose existence position does not change with time in a two-dimensional space is simulated as a wall or a pillar in a three-dimensional space including a time axis t. In addition, events that occur exactly at fixed times, such as entrances and exits with traffic lights,
In the three-dimensional space including the time axis t, the window is simulated.
In addition, a moving means that is operated at regular intervals, such as a train, is simulated as a tunnel in a three-dimensional space including the time axis t. Also, in a three-dimensional space including the time axis t, a moving means that moves at a constant speed that can be used at any time, such as a moving sidewalk, is simulated as a corridor or a tunnel with a restricted moving direction. This is to search for a route of a mobile robot.

[0034]

Therefore, accurate future prediction is impossible, and if there is uncertainty in those predictions, the uncertainty must be treated as an occurrence probability distribution. In that case, the temporal and spatial extent of the event
Depending on the ambiguity of the prediction, it will be widespread. Therefore, it is difficult to completely achieve or avoid encountering the event.

In particular, by excluding from the route search space those areas where there is little possibility that an event will occur, the movable area may become very narrow or nonexistent. For example, as shown in FIG. 18, consider a situation in which the vehicle is sandwiched between moving objects that can be stochastically predicted in an environment where the moving direction is restricted, such as a road or a corridor. In this case, it can be treated as a tunnel that becomes narrower as it goes earlier, and a route can be planned. However, as shown in FIG. If it is blocked, the front and rear moving objects must be treated as walls, and there is a problem that a route that passes between them cannot be generated.

However, in an actual scene, the path prediction of the moving object is clarified by the moment-to-moment sensing, and the distribution range is narrowed, or when there is effective means for avoiding collision with the moving object, for example, In the case where a detour path exists, or the damage due to collision is extremely small, it is possible to move to an area where the existence probability of the moving object is small to some extent while maintaining safety. Taking such a moving route is also a request from the viewpoint of the efficiency of the route search.

As described above, it is necessary for the route search to take different responses to the distribution of the occurrence probability of an event depending on the target event or situation. SUMMARY OF THE INVENTION In view of the above-mentioned conventional disadvantages, the present invention incorporates knowledge of a time-varying event into a route search in a route search device of a mobile robot that incorporates a time axis into a search space, and accurately determines the possibility of reaching the event. It is an object of the present invention to provide a route search device capable of determining different distributions of occurrence probability according to a target event or situation and searching for a route with the minimum cost.

[0038]

FIGS. 1 to 5 show an embodiment of the present invention. FIG. 1 shows an example of the configuration of a route search device, and FIG. 2 shows two moving objects. FIG. 3 shows an existence probability distribution diagram of a state in which the existence probability distribution of
FIG. 4 shows a cost distribution diagram in the case where the moving object is sandwiched between
Fig. 5 shows a cost distribution diagram in the case where the area below the threshold is cut off and sandwiched between two moving objects, and Fig. 5 shows a case where the area below the threshold is cut and the cost of the remaining area is set to a constant value. FIG. 5 shows a cost distribution diagram in the case of being sandwiched between two moving bodies. The above problem is solved by a route search device that incorporates a time axis configured as described below into a search space.

(1) In a two-dimensional or three-dimensional search space used for searching for a route of a mobile robot in which a map or information of a moving object is described, some route candidates are found. A route search device that determines a route with the minimum cost and searches for a route on which the mobile robot moves, in a search space obtained by adding a time axis to the two-dimensional or three-dimensional search space axis, A database consisting of a simulation data storage unit 1a that stores the map, the moving object time table, the motion law of the moving object, and the like, in order to process time-varying events by replacing them with events on the spatial axis. Using unit 1 and the simulation data stored in the database unit 1, a map including a probability distribution predicted value for the mobile robot or a cubic grid is generated. And that the simulation unit 2,
A grid with probability distribution predictions for the mobile robot,
Or a map consisting of a cubic grid, a cost grid, or
A cost calculation unit 3 for converting into a cubic grid map, and a route search unit for finding some route candidates from the above-mentioned cost grid or cubic grid map and determining a moving route with the minimum cost. 4 and so on.

(2) A two-dimensional or three-dimensional search space used for searching for a route of a mobile robot, in which a map or information on a moving object is described, finds some route candidates, A route search device that determines a route with the minimum cost and searches for a route on which the mobile robot moves, in a search space obtained by adding a time axis to the two-dimensional or three-dimensional search space axis, A simulation data storage unit 1a storing the map, the time table of the moving object, the law of motion of the moving object, and the like, in order to process the time-varying event by replacing it with the event on the spatial axis. A cost coefficient storage unit (1) that stores, for each event in the search space, a coefficient that varies depending on the quality and quantity of the influence on the route search.
b) a database section consisting of
Using the simulation data stored in 1, a simulation unit 2 that generates a map having a probability distribution predicted value for the mobile robot or a cubic grid map, and a probability distribution predicted value for the mobile robot. When converting a map with a grid or a cubic grid into a grid with a cost or a map with a cubic grid,
A cost calculation unit 3 that calculates a cost by multiplying the probability distribution predicted value for the mobile robot by a coefficient that varies depending on the quality and amount of influence on the route search stored in the cost coefficient storage unit 1b of the database unit 1. And a route search unit 4 that finds some route candidates from the cost grid or the cubic grid map and determines the moving route with the minimum cost.

(3) A two-dimensional or three-dimensional search space, which describes a map and information on a moving object and is used for a route search of a mobile robot, finds some route candidates. A route search device that determines a route with the minimum cost and searches for a route on which the mobile robot moves, in a search space obtained by adding a time axis to the two-dimensional or three-dimensional search space axis, A simulation data storage unit 1a storing the map, the time table of the moving object, the law of motion of the moving object, and the like, in order to process the time-varying event by replacing it with the event on the spatial axis. For each event in the search space, a coefficient that varies depending on the quality and quantity of the influence on the route search and the distribution area of the occurrence probability set for each event are cut to a predetermined value or less, and cost calculation is performed. Time threshold Using a database unit 1 consisting of a cost coefficient storage unit 1c and a simulation data stored in the database unit 1, a grid having a probability distribution predicted value for the mobile robot, or A simulation unit 2 for generating a map composed of a cubic lattice, and a lattice having a probability distribution predicted value for the mobile robot, or a map composed of a cubic lattice,
When converting to a cost grid or a map consisting of a cubic grid, the probability distribution predicted value for the mobile robot is calculated using a coefficient that varies depending on the quality and quantity of the effect on the route search stored in the database unit 1. When performing cost calculation by multiplication, the distribution area of the occurrence probability set for each of the above-mentioned events is cut out of an area that is equal to or less than a predetermined threshold stored in the cost coefficient storage unit 1c of the database unit 1 and the cost calculation is performed. The cost calculation unit 3 to be performed and the above-mentioned cost grid or
A route search unit that finds some route candidates from a map consisting of a cubic grid and determines the travel route with the lowest cost.
And it consists of.

(4) A route search device incorporating the time axis described in the item (3) into a search space, wherein the cost calculation unit 3
Then, when performing the cost calculation by cutting an area equal to or less than a predetermined threshold value stored in the cost coefficient storage unit 1c of the database unit 1, the cost calculation is performed with the cost of the uncut area as a constant value. Constitute.

[0043]

FIG. 1 shows an example of the configuration of a route search device according to the present invention. In the route search device according to the present invention shown in the figure, a database unit 1 includes a simulation data storage unit for storing knowledge for simulation of a map, a time table, a motion rule of a moving object, and the like. 1a
And a cost coefficient storage unit 1b for storing a coefficient and a threshold used for cost calculation for each event or situation.

The simulation section 2 performs a simulation for each event based on the information in the simulation data storage section 1a. At this time, interactions between events occurring in the environment of the mobile robot are not taken into account. If the event cannot be accurately predicted, the prediction is performed in the form of the distribution of the occurrence probability of the event.

The cost calculation unit 3 cuts, for each event or situation, a region where the occurrence probability of the event is equal to or less than the threshold value, and stores the cost coefficient of the cost coefficient storage unit 1b as the occurrence probability of the event. The cost distribution in the search space is calculated by performing one or both of the processes.

The route search unit 4 performs a route search by a route search method that minimizes the total cost. As described above, it is impossible to completely predict the events that will occur in the future, and the prediction is described in the form of a probability distribution. The size of the range of the distribution. Unless there is a complete barrier, trajectory, speed-limiting device, safety device, etc., the probability of occurrence of an event will not be completely “0” in any space or time.
The area of the probability distribution becomes wider and may cover the entire search space.

In order to completely eliminate the danger of encountering a certain event (collision with a moving object), setting a route by completely avoiding the distribution region of the occurrence probability of the event is as follows: Inefficient and in the worst case impossible.

In order to solve this problem, in the present invention, a path plan is made in consideration of some danger. In other words, the route planning is performed by using the occurrence probability of a certain event at the time and the place in the cost calculation when the route planning is performed, and allowing the passage through the distribution region having the occurrence probability of a certain level or less. .

On the other hand, the degree of influence on the movement of the mobile robot differs for each event. Some events are so dangerous that they should never be encountered, while others have no effect. For example, when a ship moves in the ocean, there are typhoons, tornadoes, etc.
There are large ships from the kind of yacht. It is very dangerous for a small yacht to encounter a large typhoon or the like, but there is no effect if a large ship encounters a small typhoon.

Therefore, the uniform use of the probability of occurrence of an event in the cost calculation is inefficient and creates the possibility of setting a path to encounter a very dangerous event in order to avoid harmless events. May be caused.

Therefore, in the present invention, when calculating the cost of a route, when the occurrence probability of an event is used, the quality of the effect of each event (in the above example, a small yacht encounters a large typhoon or the like) It is very dangerous, but it depends on the quality of the impact that a large ship encounters a small typhoon without any effect) and on the quantity (impact, ie how it breaks in this example) The coefficient is multiplied by the occurrence probability.

In order to simplify the above calculation, a different threshold value is provided for each event, and a distribution region where the occurrence probability of the event is equal to or less than the threshold value is cut. Further, when simplifying the cost calculation, by setting the cost in the uncut area to a constant value, the cost calculation for each route is simplified.

Since the above-described operation is performed, a path search in an environment including an event having an uncertain element in spatiotemporal space can be performed in the same manner as in an environment that does not change over time. Searching can be used. Further, the difference in the degree of influence for each event can be used for calculating the cost map used for the route search, and a path with a small degree of influence for each event can be searched.

[0054]

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 to 5 are diagrams showing an embodiment of the present invention. FIG. 1 shows an example of the configuration of a route search device of the present invention. FIG. 2 shows the existence probability of two moving objects. FIG. 3 shows an existence probability distribution diagram in a state where the distribution is blocking the course, FIG. 3 shows a cost distribution diagram in a case where the distribution is sandwiched between two moving objects, and FIG. FIG. 5 shows a cost distribution diagram in a case where the moving object is sandwiched by two moving objects, and FIG. 5 illustrates a case where the region below the threshold is cut and the cost of the remaining region is set to a constant value. The cost distribution diagram in the case is shown.

In the present invention, a map or information of a moving object is described, and several route candidates are found in a two-dimensional or three-dimensional search space used for a route search of a mobile robot. In order to determine the route with the minimum cost and search for the route on which the mobile robot moves,
Simulation data storage unit that stores maps, timetables of moving objects, laws of motion of moving objects, etc. 1a
And a coefficient that varies depending on the quality and quantity of influence on the route search for each event in the spatial axis, and a region where the occurrence probability distribution region set for each event is equal to or less than a predetermined value is cut to perform cost calculation. A database unit 1 comprising a cost coefficient storage unit 1b, 1c storing a threshold value for performing the simulation, and a simulation unit for generating a map composed of a grid or a cubic grid with a probability distribution predicted value for the mobile robot. 2, a cost calculator 3 for converting a grid or a cubic grid map having a probability distribution predicted value for the mobile robot into a cost grid or a cubic grid map; Alternatively, a route search unit 4 that finds some route candidates from a map formed of a cubic grid and determines a moving route with the minimum cost is a necessary means for implementing the present invention. Note that the same reference numerals indicate the same object throughout the drawings.

The configuration and operation of the route search device according to the present invention will be described below with reference to FIGS. As described above, in the route search device of the present invention shown in FIG.
The database unit 1 stores a simulation data storage unit 1a that stores knowledge for simulations such as maps, timetables, and the laws of motion of moving objects, and stores coefficients and thresholds used when calculating costs for each event or situation. And a cost coefficient storage unit 1b.

The simulation section 2 performs a simulation for each event based on the information in the simulation data storage section 1a. At this time, interactions between events occurring in the environment of the mobile robot are not taken into account. If the event cannot be accurately predicted, the prediction is performed in the form of the distribution of the occurrence probability of the event.

The cost calculation unit 3 cuts, for each event or situation, a region where the occurrence probability of the event is equal to or less than the threshold value, and stores the cost coefficient in the cost coefficient storage unit 1b as the occurrence probability of the event. The cost distribution in the search space is calculated by performing one or both of the processes.

The route search unit 4 performs a route search by a route search method that minimizes the total cost. FIG. 2 shows the existence probability distribution in a state where the existence probability distributions of the two moving objects block the course, as described above, by shading. When such a predetermined existence probability exists, if the path of the mobile robot is blocked, the path of the mobile robot is blocked at a portion where the existence probabilities of the mobile object A and the mobile object B overlap. I will.

FIG. 3 is a cost distribution diagram when costs are calculated in accordance with the existence probability distribution of FIG. 2. In this embodiment, since the moving object A is larger than the moving object B, the cost distribution of the moving object A is shown. Is larger than the cost distribution of the moving object B. FIG. 5 is a cost distribution diagram in a case where the existence probability of an area whose existence probability is smaller than a predetermined threshold is cut off and the cost of that part is set to “0”, as shown in FIG. By performing such a cost calculation, the mobile robot can search for a route that can avoid collision with the moving objects A and B.

FIG. 5 shows a cost distribution diagram in a case where the existence probability of a region where the existence probability of the moving object is smaller than a predetermined threshold is cut and the cost of the remaining region is kept constant. By performing such cost calculation, the cost distribution in the search space can be calculated at high speed, and the route search can be easily performed. In the route search using the time axis in the search space according to the present invention, when calculating the cost, the existence probability of the moving object in the search space is determined by P _it
(i indicates an area and t indicates a time), and the quality of the degree of influence of each event occurring between the moving objects A, B,. A, B, and the difference and the amount of influence depending on the size of the mobile robot, for example,
A product obtained by multiplying the moving objects A and B by different coefficients α _it depending on the amount of breakage caused by the collision of the mobile robot, that is, P _it × α _it is defined as the cost C _it at the space i and the time t. A search is made for a route that minimizes the set of costs ΣP _it × α _it on the route traveled by the mobile robot.

The factors that determine the above-mentioned cost coefficient α _it include, in addition to the size of the above-mentioned moving body, the weight _, speed _, material _, presence / absence of _a shock absorbing device of the moving body, its performance _, fragility, and valuableness. Degree _, ease of discovery by sensing from the mobile robot, collision avoidance capability of the moving body itself, etc. are improved.

At this time, the existence probability P _it of the moving object in the search space in the search space is equal to or less than a predetermined threshold.
By cutting the _it, it performs efficient cost calculation, to perform the route search.

Further, by setting the cost C _it of the remaining area, which has been cut by the cost calculation as described above, to a predetermined constant value {see FIG. 5}, each cost in the search space and the cubic grid is calculated. Can be easily calculated, and the calculation of the moving cost of the mobile robot can be speeded up.

In the above-described cost calculation, an example of calculation using an absolute value has been described. However, it is needless to say that calculation may be performed using a normalized cost. As described above, the route search apparatus according to the present invention that incorporates the time axis into the search space describes a map or information of a moving object, and is used for a two-dimensional or three-dimensional route used for a mobile robot's route search. A route search device that performs a route search by replacing a time-varying event in a search space to which a time axis is added to a search space axis with an event in the space axis. When calculating the cost of an event element that is only stochastically described and replaced by an event, the occurrence probability of the uncertain element is used. At this time, a different coefficient is multiplied depending on the quality and quantity of the influence on the route search for each event on the space axis. Further, the cost calculation is performed by cutting an area where the occurrence probability distribution area set for each event is equal to or less than a predetermined value.
Another feature is that the cost of the uncut area is set to a constant value in the above cost calculation.

[0066]

As described in detail above, according to the present invention, a route search in an environment including an event having an uncertain element in spatiotemporal space can be performed in an environment that does not change over time. It is possible to use a route search by cost minimization similar to the above. Further, the difference in the degree of influence for each event can be used for calculating the cost map used for the route search, and a route with a small degree of influence can be searched for each event.

[Brief description of the drawings]

FIG. 1 shows an embodiment of the present invention (part 1).

FIG. 2 shows an embodiment of the present invention (part 2).

FIG. 3 shows an embodiment of the present invention (part 3).

FIG. 4 shows an embodiment of the present invention (part 4).

FIG. 5 shows an embodiment of the present invention (part 5).

FIG. 6 is a diagram showing a configuration example of a conventional mobile robot.

FIG. 7 is a diagram showing an example of event handling in a search space incorporating a time axis (part 1)

FIG. 8 is a diagram showing an example of handling of events in a search space incorporating a time axis (part 2)

FIG. 9 is a diagram showing an example of event handling in a search space incorporating a time axis (part 3);

FIG. 10 is a diagram showing an example of event handling in a search space incorporating a time axis (part 4);

FIG. 11 is a diagram showing an example of event handling in a search space incorporating a time axis (part 5);

FIG. 12 is a diagram showing an example of event handling in a search space incorporating a time axis (part 6);

FIG. 13 is a view showing an example of event handling in a search space incorporating a time axis (part 7);

FIG. 14 is a diagram illustrating an example of handling of an event in a search space incorporating a time axis (part 8);

FIG. 15 is a view showing an example of handling of an event in a search space incorporating a time axis (part 9);

FIG. 16 is a diagram showing an example of handling of events in a search space incorporating a time axis (part 10).

FIG. 17 is a diagram showing an example of handling of events in a search space incorporating a time axis (part 11).

FIG. 18 is a diagram showing an example of event handling in a search space incorporating a time axis (part 12)

[Explanation of symbols]

 1 Database unit 1a Simulation data storage unit 1b, 1c Cost coefficient storage unit 2 Simulation unit 3 Cost calculation unit 4 Route search unit

────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-7-225612 (JP, A) (58) Field surveyed (Int. Cl. ⁷ , DB name) G05D 1/00-1/12 G06N 3 / 00

Claims (4)

(57) [Claims] 1. A two-dimensional or three-dimensional search space, which describes a map or information of a moving object and is used for a route search of a mobile robot, finds some route candidates and costs them. A route search device that determines the minimum route of the search robot and searches for a route on which the mobile robot moves, wherein the search space is obtained by adding a time axis to the two-dimensional or three-dimensional search space axis. A data storage unit for simulation that stores the map, the timetable of the moving object, the law of motion of the moving object, and the like, in order to process the event that changes in time with the event on the spatial axis.
Using a database unit (1) composed of (1a) and simulation data stored in the database unit (1), a grid having a probability distribution predicted value for the mobile robot, or a map composed of a cubic lattice A simulation unit (2) that generates a probability distribution prediction value for the mobile robot;
Or a map consisting of a cubic grid, a cost grid, or
A cost calculation unit (3) for converting the map into a cubic grid map;
A route search unit (4) that finds some route candidates and determines a moving route with the minimum cost; and a time axis incorporated in the search space. 2. A two-dimensional or three-dimensional search space, which describes a map or information of a moving object and is used for a route search of a mobile robot, finds some route candidates and calculates cost. A route search device that determines the minimum route of the search robot and searches for a route on which the mobile robot moves, wherein the search space is obtained by adding a time axis to the two-dimensional or three-dimensional search space axis. A data storage unit for simulation that stores the map, the timetable of the moving object, the law of motion of the moving object, and the like, in order to process the event that changes in time with the event on the spatial axis.
A database unit (1) comprising (1a) and a cost coefficient storage unit (1b) storing coefficients different depending on the quality and quantity of influence on the route search for each event in the search space; and A simulation unit (2) for generating a map comprising a probability distribution predicted value for the mobile robot or a cubic grid using the simulation data stored in the unit (1); Grid with probability distribution predictions,
Or a map consisting of a cubic grid, a cost grid, or
When converting to a map consisting of a cubic grid, the quality and quantity of the effect on the route search stored in the cost coefficient storage unit (1b) of the database unit (1) are added to the probability distribution predicted value for the mobile robot. A cost calculation unit (3) that performs cost calculation by multiplying by a coefficient different from the above, and a map of the above cost grid or cubic grid,
A route search unit (4) that finds some route candidates and determines a moving route with the minimum cost; and a time axis incorporated in the search space. 3. A two-dimensional or three-dimensional search space, which describes a map or information of a moving object and is used for a route search of a mobile robot, finds some route candidates and calculates a cost. A route search device that determines the minimum route of the search robot and searches for a route on which the mobile robot moves, wherein the search space is obtained by adding a time axis to the two-dimensional or three-dimensional search space axis. A data storage unit for simulation that stores the map, the timetable of the moving object, the law of motion of the moving object, and the like, in order to process the event that changes in time with the event on the spatial axis.
(1a), for each event in the search space, a coefficient that varies depending on the quality and quantity of the influence on the route search, and a region where the occurrence probability distribution region set for each event is less than or equal to a predetermined value is cut. A database unit (1) comprising a cost coefficient storage unit (1c) storing a threshold when performing cost calculation
A simulation unit (2) that generates a map comprising a probability distribution predicted value for the mobile robot or a cubic grid using the simulation data stored in the database unit (1); A grid with probability distribution predictions for the mobile robot,
Or a map consisting of a cubic grid, a cost grid, or
When converting to a map consisting of a cubic grid, cost calculation is performed by multiplying the predicted value of the probability distribution for the mobile robot by a coefficient that depends on the quality and quantity of the effect on the route search stored in the database (1). When performing, the distribution of the occurrence probability set for each of the above-mentioned events is cut out of a region equal to or less than a predetermined threshold value stored in the cost coefficient storage unit (1c) of the database unit (1), and the cost calculation is performed. From the cost calculation unit (3) to be performed and the above-mentioned cost grid or cubic grid map,
A route search unit (4) that finds some route candidates and determines a moving route with the minimum cost; and a time axis incorporated in the search space. 4. A route search device incorporating a time axis according to claim 3 into a search space, wherein said cost calculation unit (3).
Then, when performing the cost calculation by cutting the area below the predetermined threshold value stored in the cost coefficient storage unit (1c) of the database unit (1), the cost calculation is performed with the cost of the uncut area as a constant value. A route search device that incorporates a time axis into a search space, characterized by performing a search.