Path generation device and method

Abstract

The purpose of the present invention is to generate movement paths for a mobile body using not only information acquired from sensors but also information that indicates the possibility of collisions with obstacles to make it possible for the mobile body to avoid obstacles in a natural manner in a greater number of circumstances.　The present invention comprises a movement state inference unit that infers the movement state of a mobile body using information acquired from sensors, a collision risk determination unit that outputs determination information that is information that indicates determination results for the avoidance of obstacles by the mobile body on the basis of the movement state and collision risk information that is information that indicates the possibility of collisions between the mobile body and the obstacles, and a path generation unit that generates a movement path for the mobile body on the basis of the determination information.

Description

Route generation device and method

[0001] The present disclosure relates to a route generation device and method.

[0002] A path generation device has been proposed that generates a path for a moving object, such as an automobile, ship, or aircraft, to move autonomously. The autonomous moving object uses sensors to detect obstacles around the object. If an obstacle that may cause a collision with the object is present on the pre-planned path, the path generation device changes the path so that the object can avoid the obstacle.

[0003] For example, Patent Document 1 discloses a method in which a moving body avoids obstacles in a natural manner by performing finely tuned, natural-feeling braking based on the detection reliability (length of time the same obstacle is detected) which is the reliability of the obstacle detection results by a sensor.

[0004] JP 2012-183868 A

[0005] However, the conventional technology described in Patent Document 1 has the following problems. For example, in an environment where the detection sensitivity of a sensor is reduced, such as at night or in fog, the detection distance of an obstacle by the sensor becomes shorter. As the detection distance of an obstacle becomes shorter, the accuracy of the detection reliability decreases. Therefore, even if braking is performed based on the detection reliability, there is a limit to how naturally a moving body can avoid an obstacle.

[0006] The present disclosure has been made to solve the above-mentioned problems, and aims to enable a moving body to avoid obstacles in a natural manner in more situations by generating a movement path for the moving body using information indicating the possibility of collision with an obstacle, rather than relying solely on information obtained from sensors.

[0007] The route generation device of the present disclosure includes a movement state estimation unit that estimates the movement state of a moving body using information acquired from a sensor; a collision risk judgment unit that outputs judgment information that indicates a judgment result for the moving body to avoid an obstacle based on the movement state and collision risk information that indicates the possibility of a collision between the moving body and an obstacle; and a route generation unit that generates a movement route for the moving body based on the judgment information.

[0008] In the route generation method of the present disclosure, a movement state estimation unit estimates the movement state of a moving body using information acquired from a sensor; a collision risk judgment unit outputs judgment information, which is information indicating a judgment result for the moving body to avoid an obstacle, based on the movement state and collision risk information, which is information indicating the possibility of a collision between the moving body and an obstacle; and a route generation unit generates a movement route for the moving body based on the judgment information.

[0009] According to the present disclosure, in many situations, a moving object can avoid obstacles in a natural way.

[0010] Fig. 1 is a functional block diagram showing the configuration of a route generation device according to embodiment 1. Fig. 2 is a diagram showing the hardware configuration of the route generation device according to embodiment 1. Fig. 3 is a diagram for explaining a specific operation of a conventional route generation device. Fig. 4 is a diagram for explaining a specific operation of a conventional route generation device. Fig. 5 is a diagram for explaining a specific operation of route generation leave according to embodiment 1. Fig. 6 is a diagram for explaining a specific operation of route generation leave according to embodiment 1. Fig. 7 is a flowchart showing the operation of the route generation device according to embodiment 1.

[0011] Hereinafter, embodiments will be described with reference to the drawings. The following embodiments are merely examples, and various modifications are possible within the scope of the present invention. In the description of the embodiments and the drawings, the same elements and corresponding elements are given the same reference numerals. The description of elements given the same reference numerals will be omitted or simplified as appropriate. In the following embodiments, "unit" may be read as "circuit," "step," "procedure," or "process" as appropriate.

[0012] Embodiment 1. *** Description of Configuration *** Fig. 1 is a functional block diagram showing the configuration of a route generation device according to this embodiment. The route generation device 10 is a device that can implement a route generation method according to this embodiment. In Fig. 1, a mobile object 100 includes the route generation device 10, various sensors 11, and a control unit 21. The route generation device 10 includes a movement state estimation unit 12, a collision risk determination unit 13, a route generation unit 16, a map information storage unit 17, an incident information storage unit 18, a risk map information storage unit 19, and a movement route information storage unit 20. The collision risk determination unit 13 includes an estimation unit 14 and a determination unit 15.

[0013] The mobile body 100 is a mobile body capable of autonomous movement. For example, the mobile body 100 may be an autonomous vehicle, personal mobility, an autonomous robot, a ship, a train, an airplane, a drone, or the like. The mobile body 100 also includes various sensors 11 necessary for autonomous movement. For example, the various sensors 11 include devices such as a Global Positioning System (GPS), an Inertial Measurement Unit (IMU), a Light Detection and Ranging (LiDAR), a millimeter-wave radar, an ultrasonic sonar, a camera, and a beacon. Note that the various sensors 11 do not necessarily include all of these devices. The devices of the various sensors 11 may be appropriately selected or omitted depending on the type of the mobile body 100, the movement environment, the device cost, and the like.

[0014] It should be noted that the various sensors 11 do not necessarily need to be provided in the mobile body 100. For example, a sensor (e.g., a surveillance camera, a beacon, a roadside device, etc.) provided outside the mobile body 100 may acquire information necessary for the autonomous movement of the mobile body 100. In other words, if the information necessary for the autonomous movement of the mobile body 100 can be provided to the mobile body 100 from outside, the various sensors 11 of the mobile body 100 may be omitted.

[0015] The moving body 100 includes a power source and a power unit required for movement, and a driven unit driven by the power unit. For example, the power source is battery power or fuel. For example, the power unit is a motor or an engine. For example, the driven unit is a wheel, a propeller, a screw, etc. The driven unit may include a steering unit for changing direction or braking the moving body 100. Note that the power source, power unit, driven unit, and steering unit are not shown in FIG. 1 .

[0016] The various sensors 11 acquire mobile object information D1 that indicates information about the state of the mobile object 100 itself. For example, the mobile object information D1 is information about the moving speed, moving direction, and posture of the mobile object 100.

[0017] The various sensors 11 also acquire obstacle information D2 indicating information about obstacles present around the mobile body 100. For example, the obstacle information D2 is information such as the type, position, size, shape, and color of the obstacle, and whether the obstacle is moving. An obstacle is an object or state that impedes the movement of the mobile body 100. For example, if the mobile body 100 is an autonomous vehicle, the obstacle may be a fallen object, a hole, a puddle, another mobile body, a person, an animal, etc. For example, if the mobile body 100 is a ship, the obstacle may be a drifting object, a reef, an animal, another mobile body, etc. For example, if the mobile body 100 is an aircraft, the obstacle may be turbulence, a thundercloud, an animal, another mobile body, etc. Note that the obstacle information D2 may be acquired by various sensors provided in external infrastructure devices such as roadside devices, surveillance cameras, beacons, and radar, or other mobile bodies. Obstacle information D2 acquired by an external infrastructure device or another mobile body may be provided to the route generation device 10 of the mobile body 100 via a wireless communication device (not shown).

[0018] The map information storage unit 17 is a storage unit that stores map information D3 that indicates the environment of the travel route of the mobile object 100. For example, the map information D3 may be information on the condition of the ground or space including roads (paved, unpaved, road slope, etc.), railroad tracks, waterways, intersections, overpasses, the number and width of lanes, road markings, signs, and traffic lights. Note that the map information D3 may also include information on the type, position, width, size, etc. of static features that may hinder the travel of the mobile object 100.

[0019] The movement state estimation unit 12 estimates the movement state of the moving body 100 using the moving body information D1 and the map information D3. The estimated movement state is output as movement state information D4. For example, the movement state includes the current position, speed, acceleration, movement direction, and posture of the moving body 100 on the map.

[0020] The incident information storage unit 18 is a storage unit that stores incident information D5 that indicates information about situations in which accidents such as collisions or near-accidents (so-called near misses) have occurred in the past. For example, the incident information D5 may include the date and time of the accident, the coordinates of the location, information about the surrounding area of ​​the location, the weather, the type of vehicle involved in the accident, and information about obstacles. The incident information D5 may be predetermined before the vehicle 100 moves, or may be changed in real time while the vehicle 100 is moving by an information update operation from outside the vehicle 100.

[0021] The risk map information storage unit 19 is a storage unit that stores risk map information D6 that indicates information regarding the potential level of danger around the mobile object 100. More specifically, the risk map information D6 can be information that indicates, on a map, the frequency of occurrence of the risk of a collision between the mobile object and an obstacle and the impact of that risk, depending on, for example, the blind spots of the mobile object 100, the positions of features around the mobile object 100, and the shapes of the features. Note that the risk map information D6 may be determined in advance before the mobile object 100 moves, or may be changed in real time while the mobile object 100 is moving, for example, by an information update operation from outside the mobile object 100.

[0022] The travel path information storage unit 20 is a storage unit that stores travel path information D7 that indicates information about the travel path, including the travel speed and travel direction, of the moving object 100. Note that the travel path information D7 may be determined in advance before the moving object 100 moves, or may be changed as appropriate while the moving object 100 is moving, for example, by an information update operation from outside the moving object 100.

[0023] The collision risk determination unit 13 uses the obstacle information D2, the movement state emotion D4, the incident information D5, the risk map information D6, and the movement route information D7 to estimate the collision risk between the mobile body 100 and the obstacle, determine whether the mobile body 100 needs to avoid the obstacle, and output determination information D9 that is information indicating the determination result for the mobile body 100 to avoid the obstacle. Next, the internal configuration of the collision risk determination unit 13 will be described.

[0024] The estimation unit 14 uses the obstacle information D2, the incident information D5, the risk map information D6, and the travel path information D7 to estimate collision risk information D8 indicating the possibility of a collision with an obstacle if the mobile object 100 travels according to the travel path information D7. More specifically, for example, the collision risk information D8 may be a numerical value indicating the probability of the mobile object 100 colliding with an obstacle at any position on the travel path represented by the travel path information D7. For example, the collision risk information D8 may be expressed as a percentage numerical value ranging from 0% to 100%. Alternatively, the collision risk information D8 may be expressed as a normalized numerical value ranging from 0 to 1. For example, when the collision risk information D8 is expressed as a percentage numerical value, a larger numerical value indicates a higher risk of the mobile object 100 colliding with an obstacle.

[0025] The estimation unit 14 can estimate the collision risk information D8, for example, using a trained model. When the obstacle information D2, incident information D5, risk map information D6, and travel path information D7 are input to the trained model, the trained model outputs collision risk information D8, which is a response to the input information. Note that the trained model can be created using machine learning methods such as deep learning and support vector machines. The collision risk information D8 can also be estimated using a rule-based method, for example, by referencing a predetermined risk map table.

[0026] Both the incident information D5 and the risk map information D6 indicate the possibility of a collision with an obstacle (e.g., another moving object, a fallen object, etc.) on the travel path. Therefore, the estimation unit 14 may estimate the collision risk information D8 using only the incident information D5. Alternatively, the estimation unit 14 may estimate the collision risk information D8 using only the risk map information D6.

[0027] The determination unit 15 uses the movement state information D4 and the collision risk information D8 to determine whether the moving body 100 needs to avoid an obstacle, and outputs determination information D9 that is information indicating the determination result for the moving body 100 to avoid the obstacle. More specifically, for example, the determination unit 15 can compare the collision risk information D8 at the position indicated by the movement state information D4 with a predetermined threshold, and if the probability of the possibility of a collision is higher than the threshold, output determination information D9 to cause the moving body 100 to start evasive action.

[0028] To check for safety when changing lanes, the determination unit 15 can determine the congestion state around the mobile object 100 using the movement state information D4 and the collision risk information D8. For example, the congestion state may be the number of mobile objects per predetermined distance around the movement route.

[0029] The determination unit 15 may determine not only whether the moving body 100 will avoid an obstacle, but also whether other actions will be performed. For example, the determination unit 15 may determine whether the moving body 100 will perform a preparatory action to prepare for avoiding an obstacle. The preparatory action indicates an action to gently decelerate the moving body 100, such as lightly braking or stopping acceleration.

[0030] The determination unit 15 may determine whether the moving body 100 will perform an avoidance operation, such as changing the course of the moving body 100 to avoid an obstacle, or strongly decelerating or stopping the moving body 100 to prevent the moving body 100 from colliding with an obstacle. The avoidance operation indicates an operation of strongly decelerating the moving body 100 to stop the moving body 100, such as applying a normal brake, or an operation of changing the course of the moving body 100.

[0031] Hereinafter, the configuration and operation of the route generation device 10 according to this embodiment will be described in more detail, assuming that the moving body 100 is an autonomous vehicle.

[0032] First, the acceleration that occurs when the moving body 100 decelerates or changes course to avoid an obstacle will be described, and a travel path that allows the moving body 100 to avoid the obstacle in a natural way will be specifically defined.

[0033] First, the acceleration (deceleration acceleration) that occurs in the moving body 100 due to deceleration caused by, for example, braking will be described. For example, when braking lightly, the deceleration acceleration is approximately 0.1 G. When braking normally, the deceleration acceleration is approximately 0.2 G. When braking hard, the deceleration acceleration is approximately 0.3 G. Here, G is a unit representing gravitational acceleration. Generally, when the deceleration acceleration exceeds 0.3 G, people feel uncomfortable. Therefore, it is preferable that the deceleration acceleration not exceed approximately 0.25 G, taking into account a margin.

[0034] Furthermore, when the moving body 100 changes course to avoid an obstacle, lateral acceleration occurs in the moving body 100. The strength of the lateral acceleration follows the numerical value of the deceleration acceleration described above. Hereinafter, the deceleration acceleration and the lateral acceleration will be collectively referred to as avoidance acceleration.

[0035] From the above, a travel route that allows the moving body 100 to avoid an obstacle in a natural way is a travel route that is made up of deceleration or course change, or deceleration and course change, in which the avoidance acceleration does not exceed approximately 0.25 G. Note that the value of the avoidance acceleration, 0.25 G, is an example, and the value of the avoidance acceleration may be changed as appropriate depending on, for example, the type of the moving body 100 and the attributes of the occupants aboard the moving body 100 (driver, passenger, etc.).

[0036] The determination unit 15 can use, for example, a rule-based method to determine the possibility of a collision with an obstacle using the collision risk information D8 and to determine the operation of the mobile body 100. Below, the rule-based determination method will be described using as an example a case where the collision risk information D8 is expressed as a percentage value ranging from 0% to 100%.

[0037] First, thresholds are defined to determine the possibility of a collision and to determine the operation of the moving body 100. Predetermined thresholds are designated as TH1 and TH2, and for example, the values ​​of the thresholds TH1 and TH2 are set to TH1 = 20% and TH2 = 80%, respectively. Note that the values ​​of the thresholds TH1 and TH2 are merely examples, and the values ​​of the thresholds TH1 and TH2 can be changed as appropriate depending on the distance from the moving body 100 to the obstacle, the probability that the obstacle exists, the moving state of the moving body 100, the conditions of the moving route, etc.

[0038] When the collision risk information D8 is equal to or greater than the threshold value TH2, the determination unit 15 determines that the possibility of a collision with an obstacle is high and outputs determination information D9 to instruct the mobile object 100 to perform a preparatory action (braking or changing course). When the collision risk information D8 is less than the threshold value TH2 and greater than or equal to the threshold value TH1, the determination unit 15 determines that the possibility of a collision with an obstacle is medium and outputs determination information D9 to instruct the mobile object 100 to perform a preparatory action. When the collision risk information D8 is less than the threshold value TH1, the determination unit 15 determines that the possibility of a collision with an obstacle is low and outputs determination information D9 to instruct the mobile object 100 to maintain its current course and speed. Note that the number of predetermined thresholds for determining the possibility of a collision is not limited to two. For example, when there are multiple types of preparatory actions (for example, separating actions of lightly braking and stopping acceleration), the predetermined threshold can be set to three or more levels.

[0039] Furthermore, the possibility of collision may be set as a continuous value using, for example, a function y=f(x) in which the collision risk information D8 is input as x and the possibility of collision is output as y. For example, the function f(x) can be determined by a statistical method such as regression analysis using data on past accidents or events that occurred shortly before an accident.

[0040] The determination unit 15 may also determine the possibility of a collision using a trained model. Specifically, when the collision risk information D8 is input to the trained model, the trained model estimates the possibility of a collision and determines the determination information D9. For example, machine learning such as deep learning or a support vector machine can be used to create the trained model.

[0041] The path generation unit 16 uses the movement state information D4, the movement path information D7, and the determination information D9 to generate a movement path that allows the moving body 100 to avoid obstacles in a natural manner. The generated (updated) movement path is re-output to the movement path information storage unit 20 as the movement path information D7, and is also output as control information D10.

[0042] The control unit 21 uses the control information D10 to control the power source, the power unit, the driven unit, and the steering unit of the moving body 100. Then, the moving body 100 can avoid obstacles in a natural manner by moving along the generated moving path.

[0043] As described above, the collision risk judgment unit 13 does not rely solely on information acquired from various sensors (i.e., obstacle information D2) to estimate the risk of collision between the mobile body 100 and an obstacle, and to determine how the mobile body 100 can avoid the obstacle, but also makes a comprehensive judgment using information indicating the possibility of collision with the obstacle or other mobile bodies (i.e., incident information D5, risk map information D6), thereby allowing the mobile body 100 to avoid the obstacle in a natural manner.

[0044] 2 is a diagram showing the hardware configuration of the route generation device 10 according to this embodiment. The route generation device 10 is, for example, a computer, a dedicated arithmetic device, or a device that combines a computer and a dedicated arithmetic device. The route generation device 10 includes a processor 31 that is an information processing unit, a memory 32 that is a storage unit, a storage device 33 that is a non-volatile storage unit, an interface 34, and a communication unit 35.

[0045] The processor 31 is connected to other hardware via a system bus and controls the other hardware. The processor 31 is an integrated circuit (IC) that performs processing. Specific examples of the processor 31 include a central processing unit (CPU), a digital signal processor (DSP), a graphics processing unit (GPU), or a field programmable gate array (FPGA).

[0046] The processor 31 executes programs stored in the memory 32. The programs include a route generation program according to this embodiment. The functions of the route generation device 10 are realized by the processor 31 that executes the programs. The processor 31 is an example of a processing circuit.

[0047] For example, the movement state estimation unit 12, the collision risk determination unit 13, and the route generation unit 14 are realized by the processor 31. For example, the map information storage unit 17, the incident information storage unit 18, the risk map information storage unit 19, and the movement route information storage unit 20 are realized by the memory 32.

[0048] The CPU performs processes such as program execution and data calculations, while the DSP performs digital signal processing such as arithmetic operations and data movement. For example, it is desirable to perform high-speed processing by the DSP instead of by the CPU for processes such as sensing sensor data obtained from millimeter-wave radar.

[0049] A GPU is a processor specialized for image processing. A GPU can perform high-speed image processing by processing multiple pixel data in parallel. A GPU can quickly process template matching, which is frequently used in image processing. For example, it is desirable to process sensor data obtained from a camera using a GPU. Processing the sensor data obtained from a camera using a CPU requires enormous processing time. Furthermore, a GPU can be used not only as a processor for image processing, but also to perform general-purpose calculations using its computing resources (GPGPU: General Purpose Computing on Graphics Processing Units). For example, by using a GPGPU to perform image processing using deep learning, obstacles and other moving objects can be detected with higher accuracy.

[0050] An FPGA is a processor whose logic circuit configuration can be programmed. FPGAs have the properties of both dedicated hardware arithmetic circuits and programmable software. Complex arithmetic and parallel processing can be executed at high speed by FPGAs.

[0051] The memory 32 is, for example, a volatile memory. The volatile memory can move data at high speed when the path generation device 10 is operating. For example, specific examples of the volatile memory include a RAM (Random Access Memory) and an SDRAM (Synchronous Dynamic Random Access Memory).

[0052] The storage device 33 is a nonvolatile memory that can continue to store execution programs and data even when the path generation device 10 is powered off. Specific examples of nonvolatile memory include an EEPROM (Electrically Erasable Programmable Read Only Memory), a HDD (Hard Disk Drive), an SSD (Solid State Drive), and a flash memory. For example, the nonvolatile memory may be a portable storage medium such as a Secure Digital (SD) (registered trademark) memory card, a CompactFlash (CF) (registered trademark), a NAND flash, a flexible disk, an optical disk, a compact disk, a Blu-ray (registered trademark) disk, or a Digital Versatile Disc (DVD). The route generation program according to this embodiment may be provided as a portable recording medium. The memory 32 is connected to the processor 31 via a memory interface (not shown). The memory interface centrally manages memory access from the processor and performs efficient memory access control. The memory interface is used for data transfer in the route generation device 10.

[0053] 1 are realized by a processor 31 that executes programs. Programs that realize the functions of the moving state estimation unit 12, collision risk determination unit 13, and route generation unit 16 are stored in a memory 32. The processor 31 reads these programs from the memory 32 and executes them. The memory 32 is also used to temporarily store intermediate information for each program.

[0054] Furthermore, the functions of the travel state estimation unit 12, the collision risk determination unit 13, and the route generation unit 16 may be realized by a logic circuit, which is hardware. In this case, logic circuit information is stored in the memory 32. The logic circuit information is read and executed by the processor 31.

[0055] The processor 31 may be configured with a plurality of processors. In this case, the plurality of processors may execute programs that realize the functions of the travel state estimation unit 12, the collision risk determination unit 13, and the route generation unit 16 in cooperation with each other.

[0056] The memory 32 stores map information D3, incident information D5, risk map information D6, and travel route information D7. Note that the map information D3, incident information D5, risk map information D6, and travel route information D7 may be stored in a storage device external to the route generation device 10. In other words, the map information storage unit 17, the incident information storage unit 18, the risk map information storage unit 19, and the travel route information storage unit 20 may be storage devices external to the route generation device 10.

[0057] The operation and effects of the path generation device 10 according to this embodiment will be described by comparing it with the operation of a conventional path generation device. FIGS. 3 and 4 are diagrams illustrating the specific operation of the conventional path generation device. FIGS. 5 and 6 are diagrams illustrating the specific operation of the path generation device 10 according to this embodiment. In each of FIGS. 3 to 6, for example, the environment (weather) in which the detection sensitivity of the sensor is reduced is described as fog, but this environment is not limited to fog. For example, various environments in which the detection sensitivity of the sensor is reduced are conceivable, such as rain, snow, nighttime, smoke from a fire or volcanic eruption, smog, and dust storms.

[0058] First, the specific operation of a conventional path generation device will be described using FIG. 3. In FIG. 3(a), a mobile object 100 is moving on lane L1 and is located at point P1. A store S is located on the side of lane L1 at point P3. Obstacle information D2 acquired by the mobile object 100 at point P1 indicates that an obstacle H (e.g., a parked vehicle) may be present at point P3, 100 to 120 meters away from the mobile object 100, and the detection reliability of the obstacle H is 60%. A cloud-like symbol C surrounding the obstacle H represents a simulated representation of the obstacle detection reliability, with a larger cloud-like symbol C representing a lower detection reliability. A dotted arrow R1 extending from the mobile object 100 indicates the planned travel route of the mobile object 100 within a certain time period from the current time. The length of the arrow is proportional to the speed of the mobile object 100. In FIG. 3( a), a conventional path generation device estimates, at point P1, that there is a possibility that an obstacle H is present near point P3, but determines that no obstacle avoidance action should be taken because the detection reliability is low.

[0059] In FIG. 3B, the moving body 100 moves from point P1 shown in FIG. 3A in the direction of arrow R1 and is now located at point P2. Furthermore, at the time of arriving at point P2, obstacle information D2 acquired by the moving body 100 indicates that there is a possibility that an obstacle H is present at a point 50 meters away from the moving body 100, and the detection reliability is 99%. In FIG. 3B, the moving body 100 does not take any evasive action while moving from point P1 to point P2, so the speed of the moving body 100 at point P2 is the same as the speed at point P1. Therefore, it is certain that the moving body 100 needs to avoid the obstacle at point P2. Therefore, the moving body 100 needs to apply the brakes. However, because the distance to the obstacle H is close, the moving body 100 is forced to brake suddenly or change course. Therefore, the moving body 100 cannot naturally avoid the obstacle H.

[0060] FIG. 4 shows another example of the operation of a conventional path generation device. In FIG. 4(a), the mobile body 100 is located at point P1 on lane L1, as in FIG. 3(a). Furthermore, obstacle information D2 acquired by the mobile body 100 at point P1 indicates that there is a possibility that an obstacle H is present at point P3, which is 100 to 120 meters away from the mobile body 100, and the detection reliability is 60%. The difference from the example shown in FIG. 3 is that in the example shown in FIG. 4, although the detection reliability is low at point P1, a decision is made to avoid obstacle H by stopping.

[0061] FIG. 4( b ) illustrates a case where the obstacle H is almost certainly present (e.g., detection reliability is 99% or higher) when the moving body 100 arrives at point P2. In this case, the moving body 100 has already decelerated to stop, and is able to stop just before the obstacle H. On the other hand, FIG. 4( c ) illustrates a case where the obstacle H is not present when the moving body 100 arrives at point P2. In this case, the moving body 100 unnecessarily decelerates. Furthermore, the deceleration of the moving body 100 is intended to stop the moving body 100, and the moving body 100 significantly decreases in speed. In order to quickly return the moving body 100 to its original speed, the moving body 100 needs to accelerate again, but this acceleration is sudden. In this case, the moving body 100 is also unable to avoid the obstacle H in a natural manner.

[0062] Next, a description will be given of the operation of the route generating device 10 according to this embodiment. Figures 5 and 6 are diagrams for explaining the specific operation of the route generating device 10 according to this embodiment.

[0063] In Fig. 5(a), the mobile object 100 is located at point P1, as in Fig. 3(a). Also, a store S is located on the side of the road in lane L1 at point P3. At point P1, obstacle information D2 acquired by the mobile object 100 indicates that an obstacle H may be present at point P3, 100 to 120 meters away from the mobile object 100, and the detection reliability of the obstacle H is 60%. A cloud-like symbol C drawn around the obstacle H simulates the detection reliability of the obstacle, and the larger the cloud-like symbol C, the lower the detection reliability.

[0064] In FIG. 5( a), the collision risk determination unit 13 estimates the collision risk between the mobile object 100 and the obstacle H at point P1 using obstacle information D2, movement state information D4, incident information D5, risk map information D6, and movement path information D7. Based on the estimated collision risk, the collision risk determination unit 13 determines whether the mobile object 100 needs to avoid the obstacle and the details of the preparatory action or actual action. In the case of FIG. 5( a), the estimation unit 14 references the incident information D5 and the risk map information D6 and confirms that a store S is located along the road near point P3. Furthermore, it is confirmed that vehicles (i.e., obstacle H) frequently stop on the road in front of the store S to load and unload luggage, and that several rear-end collisions have occurred there in the past. Based on these results, the estimation unit 14 estimates that the collision risk near point P3 is medium.

[0065] At point P1, the probability that an obstacle H exists (detection reliability) is medium, at 60%, and it is unclear whether the moving body 100 will definitely stop. However, as shown in FIG. 5( b), the collision risk determination unit 13 prepares for the case where it determines that the collision risk is medium and therefore that it is highly necessary to stop the moving body 100. In other words, the collision risk determination unit 13 determines that light braking (i.e., preparatory movement) is necessary in advance so that the moving body 100 can stop with ease. Then, the path generation unit 16 generates a path that slows the moving speed of the moving body 100, as indicated by arrow R3, in accordance with the determination information D9.

[0066] Next, FIG. 5( c) shows a case where the moving body 100 has traveled to point P2 and it has been almost certain that an obstacle H exists. In this case, the collision risk determination unit 13 determines that it is necessary to stop the moving body 100 before the obstacle H. Then, the path generation unit 16 further changes the moving path as indicated by arrow R4 in accordance with the determination information D9 in order to stop the moving body 100 before the obstacle H. In other words, the path generation unit 16 decelerates the moving body 100 further than the speed indicated by the moving path indicated by arrow R4, and stops the moving body 100 before the obstacle H. By gradually reducing the speed of the moving body 100, the deceleration of the moving body 100 until it reaches the obstacle H becomes gentle. Therefore, the moving body 100 can avoid (stop) the obstacle H in a natural manner.

[0067] 5(d) shows an example in which the moving body 100 has moved to point P2 and it has been determined that the obstacle H is not present. In this case, the collision risk determination unit 13 determines that the preparatory action for avoidance has ended. Then, the path generation unit 16 generates an accelerating movement path so that the moving body 100 returns to the speed defined by the movement path. In this case, the moving body 100 has only minimally decelerated, so it can return to its original speed with a slight re-acceleration. Therefore, the moving body 100 can avoid the obstacle H in a natural manner.

[0068] Next, another example will be described with reference to Fig. 6. The example shown in Fig. 6 shows a case where the collision risk determination unit 13 determines that there is a possibility that an obstacle H exists, based on the distance to the obstacle H, the detection reliability, the traveling speed of the mobile body 100, and the like, but postpones the determination until the obstacle H approaches a little closer.

[0069] In Fig. 6(a), the mobile object 100 is located at point P1, as in Fig. 5(a). Also, a store S is located on the side of the road in lane L1 at point P3. At point P1, obstacle information D2 acquired by the mobile object 100 indicates that an obstacle H may be present at point P3, 100 to 120 meters away from the mobile object 100, and the detection reliability of the obstacle H is 60%. A cloud-like symbol C drawn around the obstacle H simulates the detection reliability of the obstacle, and the larger the cloud-like symbol C, the lower the detection reliability.

[0070] FIG. 6( b) illustrates a case in which the moving body 100 moves to point P4, the distance to the obstacle H becomes closer, and the detection reliability increases from 60% to 75%. At this point, the collision risk determination unit 13 determines that a preparatory movement to avoid the obstacle is necessary. Next, the collision risk determination unit 13 determines the congestion situation around the moving body 100 (i.e., the possibility of a collision with another moving body) using the movement state information D4, incident information D5, risk map information D6, and movement path information D7. In FIG. 6( b), because there are no other moving bodies around the moving body 100, the collision risk determination unit 13 determines that the congestion level of other moving bodies around the moving body is low. Unlike the example of FIG. 5, the collision risk determination unit 13 determines that the obstacle should be avoided by changing the movement path instead of stopping the moving body 100. The route generation unit 16 generates a travel route indicated by an arrow R6 in accordance with the determination information D9 from the collision risk determination unit 13 so that the moving body 100 changes course into the adjacent lane L2.

[0071] 6C shows a case where the moving body 100 has moved to point P5 and the presence of obstacle H has been almost confirmed. In this case, the moving body 100 has already changed course to the adjacent lane L2, so it continues moving and passes by obstacle H.

[0072] 6(d) shows a case where there is no obstacle H. In this case, the collision risk determination unit 13 determines to end the avoidance operation. Then, the path generation unit 16 generates a path that returns to the previous travel path, as indicated by an arrow R7.

[0073] As described above, the path generation device 10 selects an avoidance means by taking into consideration the surrounding conditions of the moving body 100, and if an obstacle H is present, the path generation device 10 avoids the obstacle H by changing the planned movement position, and if the obstacle H is not present, the path generation device 10 can quickly return to the original (pre-change) planned movement path without interrupting the movement of surrounding moving bodies. Therefore, the moving body 100 can avoid the obstacle H in a natural way.

[0074] 5 and 6 , collision risk information D8, which is information indicating the possibility of a collision between the moving body 100 and obstacle H, is a collision risk estimated from the fact that several rear-end collisions have occurred in the past on the road in front of a store S near point P3, due to vehicles frequently stopping to load and unload luggage. The collision risk determination unit 13 outputs determination information D9, which is information indicating the determination result for the moving body 100 to avoid obstacle H, based on the movement state of the moving body 100 and the collision risk information D8. Then, the path generation unit 16 can generate a movement path for the moving body 100 based on the determination information D9. This allows the moving body 100 to avoid obstacle H in a natural manner.

[0075] As described above, in the path generation device 10 according to the present embodiment, the collision risk determination unit 13 does not rely solely on information acquired from various sensors (i.e., obstacle information D2), but also makes a comprehensive determination using information indicating the possibility of a collision with an obstacle or another moving body (i.e., incident information D5 and risk map information D6). Therefore, when the possibility of the presence of an obstacle H arises, the path generation device 10 can cause the moving body 100 to perform the minimum necessary preparatory action. As a result, the moving body 100 can reliably avoid or stop the obstacle H in a natural manner. Furthermore, even if the obstacle H is not present, the path generation device 100 can quickly return the moving body 100 to the speed defined in the travel path with only minimal deceleration. Therefore, the path generation device 10 according to the present embodiment allows the moving body 100 to avoid obstacles in a natural manner in various situations.

[0076] ***Description of Operation Sequence*** Next, a description will be given of the operation sequence of the path generation device 10 according to this embodiment. FIG.

[0077] In step ST10, the moving state estimating unit 12 estimates the moving state of the moving body 100 using the moving body information D1 and the map information D3 (step ST10).

[0078] In step ST11, the estimation unit 14 uses the obstacle information D2, incident information D5, risk map information D6, and travel route information D7 to estimate collision risk information D8 indicating the possibility of a collision with an obstacle if the moving body 100 moves according to the travel route information D7 (step ST11).

[0079] In step ST12, the judgment unit 15 uses the movement status information D4 and the collision risk information D8 to judge whether the moving body 100 needs to avoid an obstacle, and outputs judgment information D9 indicating the avoidance judgment result (step ST12).

[0080] In step ST13, the path generation unit 16 generates the path information D7 based on the determination information D9 using the movement state information D4, the movement path information D7, and the determination information D9 (step ST13). Specifically, based on the determination information D9, the path generation unit 16 generates a path along which the moving body 100 performs a preparatory action to avoid the obstacle, such as light braking (deceleration), a path along which the moving body 100 performs a main action to avoid the obstacle, such as braking (deceleration) or a course change, or a path along which the moving body 100 maintains its current course and movement speed.

[0081] In step ST14, it is determined whether the moving body 100 has arrived at the destination. If the moving body 100 has arrived at the destination (Yes in step ST14), this processing flow ends. If the moving body 100 has not arrived (No in step ST14), the processing returns to step ST10. Then, the above-mentioned estimation of the moving state of the moving body 100, estimation of the collision risk with an obstacle, output of determination information D9, and generation of the moving path of the moving body 100 are each executed.

[0082] Effect of First Embodiment As described above, according to the path generation device of this embodiment, the collision risk determination unit 13 estimates the risk of collision between the mobile object 100 and an obstacle and determines how the mobile object 100 will avoid the obstacle, not only by relying on information acquired from the various sensors 11 (i.e., obstacle information D2), but also by using information indicating the possibility of collision with the obstacle or another mobile object (i.e., incident information D5 and risk map information D6), and outputs determination information D9. The path generation unit 16 then generates a travel path based on the determination information D9. As a result, the mobile object 100 can quickly determine the possibility of collision with an obstacle and travel along a travel path that allows appropriate avoidance operations. This allows the mobile object 100 to avoid obstacles in a natural manner.

[0083] In the above-described embodiment, an autonomous vehicle has been described as an example of a moving body, but the moving body is not limited to an autonomous vehicle. For example, the moving body may be a train, a ship, an airplane, etc. For example, if the moving body is a train, the route generation device according to the present disclosure can generate a moving path by regarding the track as the moving path. Furthermore, for example, if the moving body is a ship, the route generation device according to the present disclosure can generate a moving path by regarding a waterway or a sea route as the moving path.

[0084] The route generation device according to the present disclosure is applicable to various moving bodies, such as automobiles, personal mobility, trains, ships, aircraft, drones, etc. In particular, the route generation device according to the present disclosure is suitable for use in autonomously moving self-driving vehicles.

[0085] 10 Route generation device, 11 Various sensors, 12 Travel state estimation unit, 13 Collision risk judgment unit, 14 Estimation unit, 15 Judgment unit, 16 Route generation unit, 17 Map information storage unit, 18 Incident information storage unit, 19 Risk map information storage unit, 20 Travel route information storage unit, 21 Control unit, 31 Processor, 32 Memory, 33 Storage device, 34 Interface, 35 Communication unit, 100 Mobile body.

Claims

1. A path generation device comprising: a movement state estimation unit that estimates a movement state of a moving body using information acquired from a sensor; a collision risk determination unit that outputs determination information, which is information indicating a determination result for the moving body to avoid an obstacle, based on the movement state and collision risk information, which is information indicating a possibility of collision between the moving body and an obstacle; and a path generation unit that generates a movement path of the moving body based on the determination information.

2. The path generation device according to claim 1, wherein the collision risk determination unit determines a preliminary operation or an avoidance operation for the moving body to avoid an obstacle based on the movement state and the collision risk information, and the path generation unit generates a movement path based on the preliminary operation or the avoidance operation.

3. The path generation device according to claim 1 or 2, wherein the collision risk information includes at least one of incident information indicating information on a situation where a past accident or an event immediately before an accident occurred, and risk map information indicating information on a potential risk level around the moving body.

4. A path generation method, comprising: the movement state estimation unit estimating a movement state of a moving body using information acquired from a sensor; the collision risk determination unit outputting determination information, which is information indicating a determination result for the moving body to avoid an obstacle, based on the movement state and collision risk information, which is information indicating a possibility of collision between the moving body and an obstacle; and the path generation unit generating a movement path of the moving body based on the determination information.

Images