JP5278137B2 - Route search device, route search method, and autonomous mobile body - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To efficiently obtain route information in performing route search using a parallel arithmetic operation. <P>SOLUTION: A route searching device 1 includes a parallel arithmetic part 112 and a controller part 111. The parallel arithmetic part 112 calculates the evaluation value of a region by a parallel arithmetic operation. A parallel arithmetic unit 112_n includes: a region selection part 114 for selecting one selection region from regions existing in the neighborhood based on the evaluation value of each region; a route direction information generation part 115 for generating route direction information showing the direction to the selection region from an own region; a reception part 116 for receiving the route direction information; and a transmission part 118 for transmitting the route direction information to the parallel arithmetic unit 112_n of the selection region, and for transmitting the generated route direction information. The controller part 111 starts the transmission of the route direction information first from the parallel arithmetic unit 112_n at the end point of movement, and obtains the system of the route direction information propagating in the parallel arithmetic part 112 from the parallel arithmetic unit 112_n at the start point of movement. <P>COPYRIGHT: (C)2011,JPO&amp;INPIT

Description

  The present invention relates to a route search device, a route search method, and an autonomous mobile body, and in particular, when performing route search using parallel operations, a route search device, a route search method, and a route search device that can efficiently acquire searched route information. And an autonomous mobile body.

  In recent years, autonomous moving bodies that search for a route from a start point to an end point and autonomously move along the searched route have been developed. As a method for searching for a route, a method using an A * (Aster) algorithm or a Dijkstra method is well known.

  Such a route search method is usually realized by using a Neumann computer. The Neumann computer has a memory and a calculation unit as independent units, and performs processing while storing a calculation process related to path search on the memory. That is, when the arithmetic unit accesses the memory, it is necessary to sequentially communicate processing data related to the route search process with the memory.

  On the other hand, compared to the case of using a Neumann computer, a method for performing a route search at high speed by performing parallel computation using a plurality of arithmetic units connected to each other has been proposed (for example, Patent Documents). 1 to 3).

  Patent Document 1 discloses a traveling salesman problem solution processing system using a parallel computer. In the solution processing system disclosed in Patent Literature 1, each cell associated with a city to be visited adds message information when receiving a message, and transmits a message to an adjacent cell via a network. It has. Also, when the start cell that transmits the message first detects the message that the received message has returned to the start cell first through all the cities, it is assumed that the message is the message that has passed through the shortest path. Message monitoring means for terminating the search process is provided.

  Patent Document 2 discloses a traveling salesman problem processing apparatus. In Patent Document 2, in a traveling salesman problem processing apparatus using parallel computation, an overall problem processing means for dividing and assigning city coordinates according to the number of cells, and a partial problem in which hierarchical city data is assigned to each cell are processed. Sub-problem processing means for obtaining a solution is provided.

  Patent Document 3 discloses a route search system. The route search system disclosed in Patent Document 4 includes an area dividing unit that divides a moving area into a plurality of areas, and a plurality of evaluation value calculators that calculate the evaluation values of the areas corresponding to the divided areas. And a route determination unit that determines a route based on the calculated evaluation value. The evaluation value calculation unit includes an evaluation value processing unit that calculates an evaluation value based on an evaluation value of a region existing in the vicinity and a movement cost from the vicinity region to the own region.

JP-A-8-278889 JP-A-8-286920 JP 2009-19922 A

  However, when a route search is performed using a Neumann computer that performs sequential processing, the arithmetic unit needs to frequently access the main memory. There is a problem of consuming a lot of calculation cost for reference. Since the von Neumann computer has an independent unit configuration of a memory and an arithmetic unit, access between the two occurs. Therefore, when the access speed to the main memory is slow, when the access frequency to the main memory increases, or when the amount of data held in the main memory increases, the calculation speed of the computer processor itself As a result, the memory access speed becomes lower than that of the memory access speed, and the memory access speed becomes a constraint condition of the processing capability related to the entire route search processing.

  Further, the techniques disclosed in Patent Documents 1 and 2 are directed to the traveling salesman problem of searching for the shortest route among routes that make a round of all cities. For this reason, using the route search methods disclosed in Patent Documents 1 and 2, it is not possible to search for an optimum route between any two points. Further, for example, in the traveling salesman problem solution processing system disclosed in Patent Document 1, each cell holds a history of route search to the cell and transmits the history information held as a message to a neighboring cell. For this reason, there is a problem that the amount of messages to be held and transmitted increases as the number of cities to be visited increases, and the processing of the entire system is delayed.

  In the technique disclosed in Patent Document 3, the evaluation value processing unit of each region calculates an evaluation value, so that an optimum route can be searched at high speed even when the route search space is very large. However, in the technique disclosed in Patent Document 3, it is possible to calculate the evaluation value of each region at high speed in the process related to the route search, but the route information determined based on the calculated evaluation value is There is a problem that it cannot be acquired efficiently. Specifically, in the technique disclosed in Patent Document 3, after calculating the evaluation value of each region by each evaluation value calculation unit that performs parallel calculation, the region starts from the region including the movement start point, An area having the smallest evaluation value is selected as a route (see paragraph 0059 and the like). Here, in order to acquire route information that is a series of selected regions, it is necessary to sequentially access all the regions on the route. Since it is necessary to perform selection processing, route information cannot be acquired efficiently.

  Therefore, according to the present invention, in a technique for performing a route search using a parallel operation unit composed of a plurality of parallel operation units connected to each other, an upper controller unit that supervises the parallel operation unit efficiently converts the searched route information. An object of the present invention is to provide a route search device, a route search method, and an autonomous mobile body that can be acquired in a systematic manner.

A route search device according to the present invention is a route search device that searches for a route from a first point existing in a route search region to a second point, and the route search region is divided into a plurality of regions. A parallel operation unit including a plurality of parallel operation units corresponding to the divided areas, and a controller unit that controls the parallel operation unit, wherein the parallel operation unit includes the plurality of parallel operation units. Calculates the evaluation value of each corresponding region by performing parallel calculation, and the parallel computing unit calculates one of the regions existing in the vicinity based on the evaluation value of each region calculated by the parallel calculation. Area selection means for selecting a selection area, path direction information generation means for generating path direction information indicating a direction from the area corresponding to itself to the selection area, and path direction information transmitted from the other parallel computing units Received And transmitting the route direction information generated by the route direction information generating unit to the parallel computing unit corresponding to the selected area after transmitting the route direction information received by the receiving unit. Transmitting means, and the controller unit includes:
After first starting transmission of the route direction information from the second parallel computing unit corresponding to the region including the second point, the sequence of the route direction information propagated in the parallel computing unit is the first The path is searched by obtaining from the first parallel computing unit corresponding to the region including the point.

  As described above, the path direction information is locally transmitted and received between the parallel computing units, so that the path direction is directed from the parallel computing unit at the second point to the parallel computing unit at the first point in the parallel computing unit. Since the information can be propagated, the host controller unit can obtain a sequence of route direction information that is a route by accessing only the parallel computing unit at the first point. Therefore, the searched route information can be efficiently acquired when the route search is performed using the parallel operation.

  The parallel computing unit includes: a reception buffer that holds path direction information transmitted from the other parallel computing units received by the receiving unit; and the path direction information or the path direction information generated in the reception buffer. A transmission buffer for holding the route direction information generated by the transmission unit, and the transmission unit receives the route direction information transmitted from the other parallel computing unit. The path direction information held in the buffer is transmitted to the parallel computing unit corresponding to the selected area via the transmission buffer, and the path direction information transmitted from the other parallel computing unit is not received for a predetermined period. The route direction information generated by the route direction information generation means is transmitted to the parallel computing unit corresponding to the selected region via the transmission buffer. It may be. In this way, each parallel computing unit transfers the route direction information received from other parallel computing units and transmits the route direction information generated by itself, so that the entire sequence of route direction information is held in the reception buffer. Instead, it is only necessary to hold the received route direction information.

  Furthermore, the region selecting means selects one selected region in a direction closer to the region including the first point from the region existing in the vicinity based on the evaluation value of each region calculated by the parallel calculation. You may make it select.

  The parallel computing unit may further include an evaluation value processing unit that calculates an evaluation value based on an evaluation value of a region existing in the vicinity and a movement cost from the vicinity region to a region corresponding to itself. Good.

  In the route search method according to the present invention, a route search region is divided into a plurality of regions, and a parallel operation unit including a plurality of parallel operation units corresponding to the divided regions and the parallel operation unit are integrated. A route search method for searching for a route from a first point existing in the route search region to a second point, wherein the parallel operation unit includes the plurality of parallel calculation units A parallel computing unit calculates an evaluation value of each corresponding region, and the parallel computing unit calculates a region existing in the vicinity based on the evaluation value of each region calculated by the parallel computing A region selection step for selecting one selection region from, a route direction information generation step for generating route direction information indicating a direction from the region corresponding to itself to the selection region, and another parallel computing unit. A receiving step of receiving route direction information; and a transmitting step of transmitting the generated route direction information to the parallel computing unit corresponding to the selected region after transmitting the received route direction information; And the controller unit first transmits the path direction information from the second parallel computing unit corresponding to the region including the second point, and then propagates in the parallel computing unit. The route is searched by acquiring a series of direction information from a first parallel computing unit corresponding to a region including the first point.

  An autonomous mobile body according to the present invention is an autonomous mobile body that searches for a route from a first point existing in a route search region to a second point, and the route search region is divided into a plurality of regions. A parallel operation unit composed of a plurality of parallel operation units corresponding to each of the divided areas, and a controller unit that supervises the parallel operation unit, the parallel operation unit of the region existing in the vicinity Based on the evaluation value and the movement cost from the neighboring region to the region corresponding to itself, the evaluation value of each corresponding region is calculated, and the parallel computing unit calculates each parallel region by calculating the parallel operation. Based on the evaluation value, a region selection unit that selects one selected region having the smallest evaluation value among the regions existing in the vicinity, and path direction information indicating the direction from the region corresponding to itself to the selected region are generated. Sutra The direction information generating unit, the receiving unit that receives the route direction information transmitted from the other parallel computing unit, and the path direction information received by the receiving unit with respect to the parallel computing unit corresponding to the selected region A transmission unit that transmits the route direction information generated by the route direction information generation unit, and the controller unit includes a second parallel computing unit corresponding to a region including the second point After the transmission of the route direction information is first started, the sequence of the route direction information propagated in the parallel computing unit is acquired from the first parallel computing unit corresponding to the region including the first point. Thus, the route is searched.

  According to the present invention, it is possible to provide a route search device, a route search method, and an autonomous mobile body that can efficiently acquire searched route information when performing a route search using parallel operations.

It is a block diagram which shows the structure of the route search apparatus which concerns on Embodiment 1 of this invention. It is a conceptual diagram which shows the some area | region which divided | segmented the route search area which concerns on Embodiment 1 of this invention. It is a conceptual diagram for demonstrating the connection relation between the discretized area | region which concerns on Embodiment 1 of this invention. It is a conceptual diagram which shows the structure of the parallel computing unit which concerns on Embodiment 1 of this invention. It is a flowchart which shows the whole structure of the route search process by the route search apparatus which concerns on Embodiment 1 of this invention. It is a flowchart which shows the calculation process of the evaluation value of the route search apparatus area | region which concerns on Embodiment 1 of this invention. It is a figure for demonstrating an example of the evaluation value calculation process by the route search apparatus which concerns on Embodiment 1 of this invention. It is a figure for demonstrating an example of the route search result by the route search apparatus which concerns on Embodiment 1 of this invention. It is a flowchart for demonstrating the process in the parallel computing unit which concerns on Embodiment 1 of this invention. It is a figure which shows an example of the data structure transmitted / received between the parallel computing units of the route search apparatus which concerns on Embodiment 1 of this invention. It is a figure for demonstrating the route direction information by the route search apparatus which concerns on Embodiment 1 of this invention. It is a figure for demonstrating the mode of propagation of the route direction information by the route search apparatus which concerns on Embodiment 1 of this invention. It is a figure for demonstrating the mode of propagation of the route direction information by the route search apparatus which concerns on Embodiment 1 of this invention.

Embodiment 1 FIG.
In the first embodiment, a case where a route search function is installed in a robot as an autonomous mobile body will be described as an example. The robot may have, for example, two wheels that are independently rotationally controlled, and more than two wheels, for example, a four-wheeled robot, a one-wheeled robot, a two-legged walk, and a four-legged walk It may be a robot. Further, the present invention is not limited to a robot, and can be applied to a two-wheeled vehicle or a four-wheeled vehicle. The robot may include a sensor or a camera that detects an obstacle, and may perform a route search so as to avoid the obstacle according to the detected obstacle.

  According to the route search function provided in the robot according to the first embodiment, in the parallel calculation unit that performs the route search process by performing parallel calculation, route direction information is locally transmitted and received between the parallel calculation units, and the second The path direction information can be propagated from the parallel computing unit at the point to the parallel computing unit at the first point. For this reason, the upper controller unit that supervises the parallel operation unit can obtain a series of route direction information that is a route by accessing only the parallel operation unit at the first point. Therefore, the searched route information can be efficiently acquired when the route search is performed using the parallel operation. In the following description, the first point will be described as the movement start point at which the robot starts moving, and the second point will be described as the movement end point, which is the movement target point. However, the route search device according to the first embodiment is arbitrary. It searches for an optimal route between two points, and the first point may be the movement end point and the second point may be the movement start point. Embodiment 1 of the present invention will be described below with reference to the drawings.

  FIG. 1 is a block diagram showing a schematic configuration of the route search apparatus according to the first embodiment. As shown in FIG. 1, the route search device 1 includes a route search processing unit 11 and a route search storage unit 12. The route search processing unit 11 includes a controller unit 111 and a parallel calculation unit 112. The parallel computing unit 112 includes a plurality of parallel computing units 112_1, 112_2,... (Hereinafter, simply referred to as a parallel computing unit 112_n). Each of the parallel computing units 112_n is a physically independent computing unit, and executes an evaluation value calculation process described later in parallel.

  In the route search storage unit 12, a grid map obtained by virtually describing grid lines connecting lattice points arranged at substantially constant intervals d (for example, 10 cm) in the shape of the entire route search region in which the robot moves. Is remembered. The robot recognizes its own position on the grid map by replacing the position information obtained from GPS or the like with its own position on this grid map. On the grid map, the location corresponding to the robot's own position, the movement end point as the destination, and the movement direction of the robot at the movement end point are specified. The robot searches for a route from the movement start point to the movement end point that is the destination at the self-position specified on the grid map, and moves according to the searched route.

  The controller unit 111 controls the parallel operation unit 112 to perform control processing, and acquires a route including a series of route direction information from the parallel operation unit 112. The details of the control processing of the parallel computing unit 112 by the controller unit 111 will be described later.

In addition, the controller unit 111 performs a dividing process of dividing the route search area in which the robot moves into a grid area. Here, it divides | segments into the several area | region surrounding the lattice point arrange | positioned at the above-mentioned substantially constant space | interval d (for example, 10 cm). Figure 2 divides the route search region is a conceptual diagram illustrating a plurality of regions C _n which is discretized. The dividing method by the controller unit 111 is not limited to this. For example, the time required for the route search can be further shortened by dividing the area into larger areas than the above-described divided areas.

And the controller part 111 determines the connection relation between the divided | segmented area | regions. Hereinafter, a region connected to the region of interest is referred to as a neighborhood region. FIG. 3 is a conceptual diagram for explaining a connection relationship between discretized regions. In FIG. 3A, adjacent regions C _{1 to} C ₈ are defined as neighboring regions for the region of interest C ₀ . Further, in FIG. 3B, in addition to the adjacent regions C _{1 to} C ₈ , the regions C _{9 to} C ₁₆ are further defined as neighboring regions with respect to the region of interest C ₀ . In this way, by making it possible to set an area that exceeds an area adjacent to the own area as a neighboring area, it is possible to create a path that passes between the own area and the neighboring areas. In this way, the angular resolution of the route from the movement start point to the movement end point can be set flexibly, and a smoother route can be created.

  The parallel computing unit 112 includes a plurality of parallel computing units 112_1, 112_2,. The plurality of parallel computing units 112_1, 112_2,... Correspond to the areas divided by the controller unit 111, respectively. FIG. 4 is a conceptual diagram showing the configuration of the parallel computing unit 112_n. As shown in FIG. 4A, the parallel computing unit 112_n is connected to at least two evaluation value storage units 32_n and evaluation value storage units 33_n, and a state setting signal line 34_n.

  The plurality of parallel computing units 112_1, 112_2,... Have the same configuration, and the route search device 1 can be configured with a simple configuration. Each parallel computing unit 112_n may be associated with each divided region on a one-to-one basis, or one parallel computing unit 112_n may correspond to a plurality of divided regions. The parallel computing unit 112_n can be easily configured by using a programmable device such as an FPGA (Field Programmable Gate Array). Then, as shown in FIG. 4B, a plurality of parallel computing units 112_n are arranged, each corresponding to a plurality of divided areas. According to such an arrangement structure, it can be easily mounted on a robot or the like.

  In response to a command from the controller unit 111, the state setting signal line 34_n transmits a state setting signal for setting the processing by the parallel computing unit 112_n to an executable state or an unexecutable state to the parallel computing unit 112_n. Thereby, the timing which performs an evaluation value calculation process etc. about the desired parallel computing unit 112_n can be adjusted. That is, when the evaluation value calculation process or the like is executed, a state setting signal enabling execution of the evaluation value calculation process is transmitted to the parallel computing unit 112_n, and when the evaluation value calculation process is not executed, the evaluation value calculation is calculated. A state setting signal that disables the process is transmitted.

  Returning to FIG. 1, the description will be continued again. The parallel computing unit 112n includes an evaluation value processing unit 113, a region selection unit 114, a path direction information generation unit 115, a reception unit 116, a reception buffer 117, a transmission unit 118, and a transmission buffer 119. Yes.

  The evaluation value processing unit 113 calculates an evaluation value based on the evaluation value of the area existing in the vicinity and the movement cost from the vicinity area to the area corresponding to itself.

  The region selection unit 114 selects one selected region from the regions existing in the vicinity based on the evaluation value of each region calculated by performing the parallel calculation. Here, the region selection unit 114 selects one selection region in a direction closer to the region including the movement start point from the regions existing in the vicinity. For example, in other areas except forbidden areas where movement is prohibited, if an evaluation value with a larger value is calculated for an area farther away from the movement start point based on the evaluation value of the area including the movement start point, the area selection The unit 114 selects one selected region having the smallest evaluation value among the regions existing in the vicinity. In the following, the direction closer to the area including the movement start point from the own area may be referred to as the upstream direction, and the direction closer to the area including the movement end point from the own area may be referred to as the downstream direction.

  The route direction information generation unit 115 generates route direction information indicating the direction from the region corresponding to itself to the selected region. The route searched by the route search device 1 is composed of a series of route direction information.

  The receiving unit 116 receives the route direction information transmitted from the other parallel computing unit 112_n. The reception buffer 117 holds the route direction information received by the reception unit 116. The data held in the reception buffer 117 is not limited to the path direction information, and may be used for receiving and holding other data transmitted / received between the parallel computing units 112_n.

  The transmitting unit 118 transmits the route direction information generated by the route direction information generating unit 115 after transmitting the route direction information received by the receiving unit 116 to the parallel computing unit 112_n corresponding to the selected region. Here, after the transmission unit 118 finishes transferring the route direction information received from the other parallel computing unit 112_n, the transmission unit 118 transmits the route direction information generated by the route direction information generation unit 115 to the transmission buffer following the transfer processing. Via 119 to the parallel computing unit 112_n corresponding to the selected region. Therefore, when the transmission unit 118 receives path direction information from another parallel computing unit 112_n, the transmission unit 118 uses the parallel computing unit corresponding to the selected region via the transmission buffer 119 to store the path direction information held in the reception buffer 117. 112_n, and when the route direction information is not received from other parallel computing units 112_n for a predetermined period, the route direction information generated by the route direction information generation unit 115 is sent to the selection region via the transmission buffer 119. It transmits to corresponding parallel computing unit 112_n.

  The transmission buffer 119 holds the route direction information held in the reception buffer 117 or the route direction information generated by the route direction information generation unit 115 for transmission. The data held in the transmission buffer 119 is not limited to the path direction information, and may be used for transmission and holding of other data transmitted / received between the parallel computing units 112_n.

  Next, route search processing by the route search device 1 according to Embodiment 1 will be described with reference to FIGS. FIG. 5 is a flowchart showing an overall configuration of route search processing by the route search device 1 according to the first embodiment.

  First, the controller unit 111 sets a movement start point and a movement end point on the grid map (step S101). Then, on the grid map, an obstacle existing in the environment is registered (step S102) and a prohibited area (step S103) is created.

  Next, the parallel computing unit 112 calculates an evaluation value of each corresponding region by performing a parallel computation by the plurality of parallel computing units 102_n (step S104). Therefore, the controller unit 111 divides the grid map into a plurality of regions, associates each region with the parallel computing unit 112_n, and then performs an evaluation value calculation process on each parallel computing unit 112_n of the lower parallel computing unit 112. Instruct the start of. Details of the evaluation value calculation process in the parallel operation unit 112 will be described later.

  Next, the controller unit 111 causes the parallel computing unit 112_n to generate route direction information of each region from the evaluation value calculated in step S104, and acquires the route direction information propagated in the parallel computing unit 112 (step S105). . The details of the process of acquiring the route direction information from the parallel computing unit 112 will be described later.

  The route search apparatus 1 according to the first embodiment enables an efficient acquisition process of route direction information according to S105 among the route search processing according to steps S101 to S105. That is, according to the route search device 1 according to the first embodiment, the route direction information can be efficiently obtained from each parallel computing unit 112_n of the parallel computing unit 112.

  First, the evaluation value calculation process according to S104 shown in FIG. 5 will be specifically described with reference to FIGS. FIG. 6 is a flowchart specifically showing the evaluation value calculation process for each region.

  First, the controller unit 111 divides the route search area into a plurality of areas, and associates the parallel computing unit 112_n with each of the divided areas (S201). Here, the route search area is divided into a number of areas equal to the number of grids. For example, when the route search area is a two-dimensional plane, a plurality of discrete areas each having an area of 10 cm × 10 cm are divided and generated.

  Next, the controller unit 111 substitutes initial values for all the parallel computing units 112_n (S202). As the state of each region, three states are assumed, which are a prohibited state in which entry is prohibited, an uncalculated state in which an evaluation value has not been calculated, and a calculated state in which an evaluation value has been calculated. An initial value corresponding to each state is assigned to the parallel computing unit 112_n. In S202, which is the start stage of the evaluation value calculation process, the evaluation value storage unit 32_n and the evaluation value storage unit 33_n of the parallel computing unit 112_n corresponding to the region other than the entry prohibition region in the route search target region are initially stored. A sufficiently large value (for example, 1.0 × 10 ^ 7) is substituted as the value. A sufficiently large value (for example, 1.0 × 10 ^ 9) is assigned as an initial value to the evaluation value storage unit 32_n and the evaluation value storage unit 33_n of the parallel computing unit 112_n corresponding to the entry prohibition region. Here, as an initial value, a value sufficiently higher than an evaluation value calculated by an evaluation value calculation process described later is adopted and substituted. The state of the region may be determined by setting such initial values, or state information indicating the state of the region may be stored in each parallel computing unit 112_n.

  Next, the controller unit 111 designates the movement start point, and for the parallel computing unit 112_n corresponding to the region including the movement start point, at least one of the evaluation value storage unit 32_n or the evaluation value storage unit 33_n A minimum value (for example, 0) is substituted (S203). In the following description, it is assumed that the minimum value of the evaluation value is assigned only to the evaluation value storage unit 32_n.

Then, the controller unit 111 generates a calculation start clock _{C 1} to initiate an evaluation value calculation processing (S204). A pulse corresponding to the clock of the CPU is transmitted to each parallel computing unit 112_n via the state setting signal line 34_n, and the processing in S205 to S208 is executed by each parallel computing unit 112_n. At this time, a state setting signal enabling execution of the evaluation value calculation process is transmitted to the parallel calculator 112_n that executes the evaluation value calculation process, and the parallel calculator 112_n that does not execute the evaluation value calculation process. Transmits a state setting signal that disables execution of the evaluation value calculation process.

  Each parallel computing unit 112_n updates the evaluation value of its own region as follows based on the evaluation value of its own region and the evaluation value of the neighboring region. First, with respect to the value stored in the evaluation value storage unit 32_n, it is determined whether or not the self-evaluation value that is the evaluation value of the self-region has been updated from the initial value (S205).

  If the result of determination in step S205 is that the self-evaluation value remains the initial value (if the evaluation value of the own region has not been calculated), the parallel calculator 112_n corresponding to the neighboring region uses the evaluation value storage unit It is determined whether or not an evaluation value that is not an initial value exists as a value stored in 32_n (S206).

If the result of determination in step S206 is that there is a value other than the initial value in the neighborhood area (if the neighborhood area evaluation value has already been calculated), the smallest assessment value among the neighborhood area assessment values and its minimum The evaluation value of the self region is calculated based on the movement cost corresponding to the geometric distance from the neighboring region having the evaluation value of to the self region, and the result is stored in the evaluation value storage unit 33_n (S207). Here, the evaluation value of its own region is calculated based on the following formula.
Self-evaluation value = minimum evaluation value of neighboring area + moving cost from neighboring area to own area

  On the other hand, as a result of the determination in step S205 or S206, when the self-evaluation value stored in the evaluation value storage unit 32_n has been updated from the initial value, or when there is no value other than the initial value in the neighborhood region, The current self-evaluation value is maintained as it is without being updated (S208). Here, as a process of maintaining the self-evaluation value, the evaluation value stored in the evaluation value storage unit 32 is output to the evaluation value storage unit 33_n and stored. By processing in this way, the current self-evaluation value can be maintained by simple processing. The evaluation value stored in the evaluation value storage unit 32_n may not be output to the evaluation value storage unit 33_n and no processing may be performed.

In all parallel computing units 112_n, execution of the evaluation value calculation processing in steps S205 to S208 is determined based on the evaluation values stored in the evaluation value storage unit 32_n, respectively, and the evaluation value calculation results are respectively determined as evaluation value storage units. The result is output to an evaluation value storage unit 33_n different from 32_n. As a result, it is possible to prevent interference between evaluation value calculation results as a result of one evaluation value calculation process, and it is possible to limit the evaluation value from being calculated beyond one vicinity region. It is possible to calculate only the evaluation value of the area to be processed. Therefore, it is possible to synchronize the timing of the evaluation value calculation processing by the plurality of parallel computing units 112_n. Incidentally, when the process of step S205 to 208 by the next clock _{C 2} clock _{C 1} is executed in the process of step S205 to S208, based on the evaluation value stored in the evaluation value storage section 33_n Rating A value calculation process is executed, and the result is stored in the evaluation value storage unit 32. Furthermore, when the process of the next clock C ₃ by S205 to 208 of the clock C ₂ is performed, the evaluation value calculation process based on the evaluation value stored in the evaluation value storage section 32_n is executed, the The result is stored in the evaluation value storage unit 33_n.

  The controller unit 111 determines whether an evaluation value that is not an initial value exists in either one of the evaluation value storage unit 32_n or the evaluation value storage unit 33_n for the parallel computing unit 112_n corresponding to the region including the movement end point. Determine (S209). If the evaluation value does not exist as a result of the determination in step S209, the evaluation value of the region including the movement end point has not yet been calculated, so the process returns to step S204 and again each parallel computing unit 112_n. Thus, the area evaluation value calculation process is executed.

  On the other hand, if the evaluation value exists at the movement end point as a result of the determination in step S209, the evaluation value calculation process is terminated because the evaluation value of the area including the movement end point has already been calculated. Then, the process proceeds to the acquisition process of the route direction information according to S105 shown in FIG.

  As a condition for terminating the evaluation value calculation process in step S209, the process may be terminated when the evaluation values corresponding to all the regions no longer change. That is, in all the parallel computing units 112_n, the processing can be terminated when the evaluation values stored in the evaluation value storage units 32_n and 33_n do not change with the same value. Further, by configuring such an end condition in addition to the end condition in step S209, the evaluation value calculation process can be ended even when there is no route from the movement start point to the movement end point. it can.

  A specific example relating to the evaluation value calculation process for each region illustrated in FIG. 6 will be described below with reference to the grid maps illustrated in FIGS. 7 and 8. Note that the grid maps shown in FIGS. 7 and 8 represent the target path serving as the target trajectory of the robot as a set of regions discretized in a grid shape. The route search storage unit 12 of the computer constituting the route search device 1 stores information about the grid map. Here, the information regarding the grid map includes coordinate information of each region, connection information between regions indicating neighboring regions of each region, and evaluation value information associated with each region. The evaluation value of the area is calculated by the parallel computing unit 112_n corresponding to each area.

  FIG. 7 is a diagram for explaining an example of the evaluation value calculation process. In FIG. 7, in order to divide the route search area into 16 horizontal areas and 13 vertical areas so that the evaluation other value calculation process can be performed easily, the neighboring areas of each area are arranged in four directions, front, rear, left, and right adjacent to each area. Adjacent area. In FIG. 8, regions located in four oblique directions from each region are also included in the neighboring region, and regions adjacent to each region in the eight directions are designated as neighboring regions. In the figure, the hatched area indicates an entry prohibition area, and the area on the outer periphery indicates an area that is not subject to route search. When the initial value of the region of the movement start point shown in the figure is set to 0, the evaluation value calculation result by the parallel computing unit 112_n is shown in each region. Here, the value of the movement cost from the neighboring area to the own area is set to 5. Of the 208 divided areas, 39 areas are entry prohibition areas and 169 areas are route search target areas. According to the route search device 1 according to the first embodiment, since each parallel computing unit 112_n calculates an evaluation value in parallel, the evaluation value can be calculated by a total of 22 calculations.

  After the evaluation value calculation processing for each region in the parallel operation unit 112 is completed, the controller unit 111 acquires the route direction information from a specific parallel computing unit 112_n in the parallel operation unit 112, thereby obtaining the route direction information series. Get configured paths efficiently. For example, in FIG. 8, the generated route direction information is indicated by white arrows (the white arrows shown between the regions indicate the route direction information generated in each region). A series of these route direction information forms a route. From the end point to the start point, the route specified along the route direction information indicates the shortest route from the end point to the start point. (At the same time, it means the shortest route from the end point to the start point.) To do). Note that the evaluation value 110 at the movement end point means the total cost of the movement cost required from the movement start point to the movement end point. As will be described later, the parallel operation unit 112 propagates the route direction information in each region toward the movement start point that is the upstream direction. The controller unit 111 sequentially acquires the path direction information from the parallel computing unit 112_n corresponding to the area including the movement start point. The controller unit 111 configures a route on the grid map (that is, a point sequence of coordinate values constituting the route) from the coordinates of the movement start point or movement end point on the grid map and the acquired series of route direction information. Thereby, the route search device 1 searches for a route from the movement start point to the movement end point.

  Next, with reference to FIGS. 9 to 13, the route direction information acquisition process according to S105 shown in FIG. 5 will be specifically described. After the evaluation value calculation process is completed, the controller unit 111 first starts transmission of route direction information from the parallel computing unit 112_n corresponding to the region including the movement end point. In the parallel computing unit 112, the route direction information of each region is propagated from the region corresponding to the movement end point toward the upstream region based on the evaluation value of each divided region. The controller unit 111 sequentially acquires the propagated route direction information from the parallel computing unit 112_n corresponding to the region including the movement start point.

  With reference to FIG. 9, the propagation process of the route direction information in the parallel operation unit 112 will be described. FIG. 9 is a flowchart for explaining the processing of each parallel computing unit 112_n. First, the reception unit 116 of the parallel computing unit 112_n receives the reception data transmitted from the parallel computing unit 112_n in the downstream direction, and holds the reception data in the reception buffer 117 (step S301). Data transmitted / received between the parallel computing units 112_n can be data having a structure as shown in FIG. For example, it comprises a mode portion indicating the mode of the route search processing and a data portion processed in the mode. For example, in the mode portion, the evaluation value calculation mode can be set during the evaluation value calculation process, and the route direction information acquisition mode can be set during the acquisition of the route direction information. In the data portion, an evaluation value of a neighborhood area and a value of route direction information are set. By configuring the data structure of the reception data in this way, only when the mode portion of the reception data indicates the route direction information acquisition mode, the following processing is performed using the route direction information set in the data portion. Good.

  Next, the parallel computing unit 112_n determines whether or not the received data is received within a predetermined period (step S302). The parallel computing unit 112_n transfers the path direction information received from the other parallel computing unit 112_n in the downstream direction to the other parallel computing unit 112_n in the upstream direction. Here, the parallel computing unit 112_n transfers all the route direction information respectively transmitted from the plurality of other parallel computing units 112_n in the downstream direction from itself, and then continues to the plurality of route direction information, as will be described later. The route direction information generated in this way is transmitted to another parallel computing unit 112_n in the upstream direction. Therefore, here, when the reception of the path direction information continuously transmitted from the other parallel computing unit 112_n in the downstream direction is interrupted (that is, when it is not received within a predetermined period), the downstream direction It is determined that the transmission of the route direction information from has ended.

  If the result of determination in step S302 is that it has been received within a predetermined period, the transmission unit 118 sets the path direction information received from other parallel computing units 112_n in the downstream direction in the transmission buffer 119 (step S303). ) To the other parallel computing unit 112_n in the upstream direction (step S304).

  As a result of the determination in step S302, if it is not received within the predetermined period, the route direction information generation unit 115 generates route direction information indicating the direction to the upstream region (step S305).

  Here, the route direction information generation processing will be described with reference to FIG. First, the region selection unit 114 selects one selection region in a direction closer to the region including the movement start point from the regions existing in the vicinity. Here, the region selection unit 114 selects one selected region having the smallest evaluation value from the regions existing in the vicinity. Then, the route direction information generation unit 115 generates information on the direction from its own region toward the selected region as route direction information. For example, as shown in FIG. 11, a white arrow heading from a central area indicating its own area to 8 neighboring areas is defined as a path direction, and these 8 directions are associated with numbers 0 to 7, respectively. For example, when the route direction information is indicated by using 8-bit information, only the value of the corresponding bit among the 8-bit values is set to 1, so that the direction of any of the neighboring 8 regions is set. Can be shown. Specifically, for example, the route direction information in the downward direction (number 4) can be indicated by “00001000”.

  In addition, although it demonstrated as what selects a selection area | region from a near area and produces | generates route direction information here during acquisition of route direction information, route direction information is produced | generated during the evaluation value calculation process mentioned above, It is good also as what is previously stored in each parallel computing unit 112_n.

  Next, the transmission unit 118 sets the route direction information generated by the route direction information generation unit 115 in the transmission buffer 119 (step S306), and transmits it to another parallel computing unit 112_n in the upstream direction (step S304). .

  12 and 13 are diagrams for explaining a state of propagation of route direction information. In FIG. 12, the area name of the movement start point is indicated as A, and the area name of the movement end point is indicated as I. The route direction information generated in each area shown in FIG. 12 propagates toward the movement start area A. Specifically, as shown in FIG. 13, first, the path direction information I (4) indicating the downward direction is generated by the parallel computing unit 112_n corresponding to the region I, and the region H selected based on the evaluation value is generated. The path direction information I (4) is transmitted to the corresponding parallel computing unit 112_n. The parallel computing unit 112_n corresponding to the region H transfers the received path direction information I (4) to the parallel computing unit 112_n corresponding to the region G in the downstream direction. Furthermore, the parallel computing unit 112_n corresponding to the region H generates path direction information H (3) indicating the lower right direction, and transmits the path direction information H (3) to the parallel computing unit 112_n corresponding to the region G. To do. Thereafter, the same processing is repeated from the region G to the region A, and the route direction information I (4), the route direction information H (3), and the route direction information G (4) for the parallel computing unit 112_n corresponding to the region A. Route direction information F (5), route direction information E (4), route direction information D (5), route direction information C (6), and route direction information B (5) are transmitted in succession.

  In this way, the route direction information generated in each region is propagated to the parallel computing unit 112_n corresponding to the region A of the movement start point. The controller unit 111 accesses the parallel computing unit 112_n corresponding to the movement start point area A, and sequentially acquires path direction information. That is, the controller unit 111 can efficiently acquire a sequence of route direction information by accessing only the parallel computing unit 112_n corresponding to the region A of the movement start point.

  Here, in the parallel operation unit 112, the calculation steps necessary for collecting all the route direction information in the parallel operation unit 112_n for the region A of the movement start point are as follows. 2n-1. When the controller unit 111 directly accesses each parallel computing unit 112_n and acquires the route direction information without propagating the route direction information in the parallel computing unit 112, the controller unit 111 accesses each area, It is necessary to perform the process of selecting the area going upstream in all areas. In contrast, in the path direction information acquisition process according to the first embodiment, the path direction information of each area is propagated to the parallel computing unit 112_n corresponding to the movement start point area in the parallel computing unit 112. Therefore, the controller unit 111 only needs to access the corresponding area and acquire the route direction information. That is, when the controller unit 111 accesses the parallel computing unit 112_n to acquire path direction information, the controller unit 111 does not need to perform a process of selecting a region from the accessed region to the upstream direction. It is possible to reduce the amount of calculation when accessing the arithmetic unit 112.

  In addition, since the controller unit 111 only needs to obtain the path direction information from the parallel computing unit 112_n corresponding to the movement start point, the number of connection lines for the upper controller unit 111 to access the lower parallel computing unit 112 is suppressed. can do. In addition, by providing one connection line for the plurality of parallel computing units 112 — n and switching the connection lines for use, the number of necessary connection lines can be further suppressed. Furthermore, since only one path direction information needs to be held in the reception buffer 117 of each parallel computing unit 112_n, the cost of the reception buffer 117 can be suppressed.

  In addition, when the size of the reception buffer 117 of the parallel computing unit 112_n corresponding to the movement start point is set to a size that can hold the entire series of route direction information constituting the route, the reception buffer mounted on each parallel computing unit 112_n. Although the cost of 117 increases, the controller unit 111 can acquire the entire series of route direction information by one access. For this reason, the amount of access between the parallel computing unit 112 and the controller unit 111 can be reduced, and the route search process can be executed more efficiently.

  As described above, the present invention relates to a method for efficiently obtaining route direction information constituting a route from each parallel computing unit 112_n in a method for calculating a route search at high speed using parallel operations. According to the present invention, the controller unit 111 can efficiently acquire a series of route direction information by using only local communication between the parallel computing units 112_n. Thereby, the communication cost between the high-order controller part 111 and the low-order parallel operation part 112 can be reduced, and the manufacturing cost and calculation cost which concern on the route search apparatus 1 can be reduced.

  In each of the above-described embodiments, the movement area in which the robot moves is described as a path search area, and the path from the movement start point to the movement end point is searched. However, the present invention is not limited to this. In other words, the present invention can be applied not only to a moving area where a robot moves, but also to a parallel operation for searching for an optimum route between any two points in a communication network or a dynamic model field. .

  In addition, this invention is not limited to each embodiment mentioned above, In the range which does not deviate from the meaning, it can change suitably.

1 route search device,
11 route search processing unit, 12 route search storage unit,
111 controller unit, 112_n parallel computing unit,
113 evaluation value processing unit, 114 area selection unit, 115 route direction information generation unit,
116 reception unit, 117 reception buffer, 118 transmission unit,
119 send buffer,
32_n evaluation value storage unit, 33_n evaluation value storage unit, 34_n state setting signal line,
C _n region

Claims (6)

A route search device for searching for a route from a first point existing in a route search region to a second point,
The route search area is divided into a plurality of areas, and a parallel operation unit composed of a plurality of parallel computing units corresponding to the divided areas,
A controller unit that supervises the parallel operation unit,
The parallel computing unit is
The plurality of parallel computing units calculate in parallel, thereby calculating an evaluation value of each corresponding region,
The parallel computing unit is:
Based on the evaluation value of each area calculated by performing the parallel operation, area selection means for selecting one selection area from the areas existing in the vicinity,
Route direction information generating means for generating route direction information indicating a direction from the region corresponding to itself to the selected region;
Receiving means for receiving path direction information transmitted from another parallel computing unit;
Transmitting means for transmitting the route direction information generated by the route direction information generating means after transmitting the route direction information received by the receiving means to the parallel computing unit corresponding to the selected area; Prepared,
The controller unit is
After first starting transmission of the route direction information from the second parallel computing unit corresponding to the region including the second point, the sequence of the route direction information propagated in the parallel computing unit is the first A route search apparatus, wherein the route is searched by obtaining from a first parallel computing unit corresponding to a region including the point. The parallel computing unit is:
A reception buffer that holds path direction information transmitted from the other parallel computing units received by the receiving unit;
A transmission buffer that holds the route direction information held in the reception buffer or the route direction information generated by the route direction information generation unit;
The transmission means includes
When path direction information transmitted from another parallel computing unit is received, the path direction information held in the reception buffer is transmitted to the parallel computing unit corresponding to the selected region via the transmission buffer. And
When the route direction information transmitted from the other parallel computing unit is not received for a predetermined period, the route direction information generated by the route direction information generating means is transferred to the selection area via the transmission buffer. The route search device according to claim 1, wherein the route search device transmits to the corresponding parallel computing unit. The region selection means includes
The one selection area in the direction closer to the area including the first point is selected from the areas existing in the vicinity based on the evaluation value of each area calculated by the parallel calculation. 1. The route search device according to 1. The parallel computing unit is:
The route according to claim 1, further comprising: an evaluation value processing unit that calculates an evaluation value based on an evaluation value of a region existing in the vicinity and a movement cost from the neighboring region to a region corresponding to itself. Search device. A routed region is divided into a plurality of regions, and a mobile unit comprising: a parallel operation unit composed of a plurality of parallel computing units corresponding to each of the divided regions; and a controller unit that supervises the parallel operation unit. A route search method for searching for a route from the first point existing in the route search region to the second point,
The parallel computing unit is
The plurality of parallel computing units calculate in parallel, thereby calculating an evaluation value of each corresponding region,
The parallel computing unit is
Based on the evaluation value of each region calculated by performing the parallel operation, a region selection step of selecting one selection region from the regions existing in the vicinity,
A route direction information generating step for generating route direction information indicating a direction from the region corresponding to itself to the selected region;
A receiving step of receiving path direction information transmitted from the other parallel computing units;
A transmission step of transmitting the generated route direction information after transmitting the received route direction information to the parallel computing unit corresponding to the selection region;
The controller unit is
After first starting transmission of the route direction information from the second parallel computing unit corresponding to the region including the second point, the sequence of the route direction information propagated in the parallel computing unit is the first A route search method, wherein the route is searched by obtaining from a first parallel computing unit corresponding to a region including the point. An autonomous mobile that searches for a route from the first point existing in the route search area to the second point,
The route search area is divided into a plurality of areas, and a parallel operation unit composed of a plurality of parallel computing units corresponding to the divided areas,
A controller unit that supervises the parallel operation unit,
The parallel computing unit is
Based on the evaluation value of the area existing in the vicinity and the movement cost from the vicinity area to the area corresponding to itself, the evaluation value of each corresponding area is calculated,
The parallel computing unit is:
Based on the evaluation value of each area calculated by performing the parallel operation, an area selection unit that selects one selection area having the smallest evaluation value among the areas existing in the vicinity;
A route direction information generating unit for generating route direction information indicating a direction from the region corresponding to itself to the selected region;
A receiving unit that receives path direction information transmitted from the other parallel computing units;
A transmission unit that transmits the route direction information generated by the route direction information generation unit after transmitting the route direction information received by the reception unit to the parallel computing unit corresponding to the selection region; Prepared,
The controller unit is
After first starting transmission of the route direction information from the second parallel computing unit corresponding to the region including the second point, the sequence of the route direction information propagated in the parallel computing unit is the first An autonomous mobile body, wherein the route is searched by obtaining from a first parallel computing unit corresponding to a region including the point.

Images