JP2015081831A - Sound source position estimation device, mobile body and control method for mobile body - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an independent type mobile body capable of controlling movement by detecting a position of another object approaching from a range obstructed from view by using reflection sound.SOLUTION: A robot 1000 includes: a steered response power acquirement unit 1050 for executing a process for acquiring spectrum for sound source power based on a signal from a microphone array 1052; arriving sound direction distribution estimation unit 1060 for estimating an arrival direction of sound on a robot coordinate based on the sound source power spectrum; a mobile body stance specification process unit 1040 for specifying the stance of the robot based on a signal from a path length measuring sensor 10 or LRF20 and 30; and a sound source position estimation process unit 1070 for estimating the sound source position based on specified robot stance, a three-dimensional geometric map 1102 and the estimation result of the arriving sound direction distribution estimation unit 1060.

Description

  The present invention relates to a sound source localization technique in a real environment, and more particularly, to a sound source position estimation technique using sound directionality by a sound sensor array in a real environment and movement control of a moving object using the sound source position.

  Different environments such as home, office, and shopping streets have various noise characteristics depending on location and time, so in applications that target specific sounds such as voice, depending on the type and degree of environmental noise used, There is a problem that the expected performance cannot be obtained.

  For example, in voice communication between a person and a robot, the microphone attached to the robot is usually at a position (1 m or more) away from the robot. Therefore, for example, the signal-to-noise ratio (SNR) is lower than when the distance between the microphone and the mouth is several centimeters as in telephone voice. For this reason, the voices of others nearby and the noise of the environment become interference sounds, making it difficult for the robot to recognize the target speech. Therefore, sound source localization and sound source separation are important for robot applications.

  Regarding the sound source localization, there are those described in Patent Document 1 or Patent Document 2 as conventional techniques assuming an actual environment. The technique described in Patent Document 1 or Patent Document 2 uses a known sound source localization method called the MUSIC method with high resolution.

  In the invention described in Patent Document 1 or Patent Document 2, a current correlation matrix is calculated based on a received signal vector obtained by Fourier-transforming a signal from the microphone array using a microphone array and a past correlation matrix. To do. The correlation matrix thus obtained is subjected to eigenvalue decomposition to obtain the maximum eigenvalue and a noise space that is an eigenvector corresponding to an eigenvalue other than the maximum eigenvalue. Furthermore, the direction of the sound source is estimated by the MUSIC method based on the phase difference of the output of each microphone, the noise space, and the maximum eigenvalue with one microphone as a reference in the microphone array.

  In most studies on sound source localization and sound source separation so far, reflected sound has been treated as having an adverse effect (for example, see Non-Patent Document 1 and Non-Patent Document 2).

  On the other hand, in order to guarantee safe movement in autonomous moving bodies such as robots, it is necessary to detect other objects approaching from areas that cannot be seen by ordinary sensors in order to ensure safe movement. Become.

  For example, if a person approaches a place where a passage intersects, if he hears someone's footsteps coming to the intersection, he will not be able to adjust his pace of movement even if he cannot see it. It is common.

  However, since the sensor mounted on the robot is generally configured to detect a signal from the line-of-sight region, it is not easy to cope with such a situation.

  In other words, the use of a laser or RGB-D camera to detect and track a moving object by ranging is clearly limited to the condition of seeing the object. In addition, an ultrasonic sensor used for robot perception in recent years evaluates a distance to a reflector by measuring a propagation time of a radiated ultrasonic pulse. The ultrasonic sensor needs to be in a position where an object that reflects a pulse can be seen through. Therefore, systems that rely on direct visibility of obstacles to use robotic localization or obstacle detection using an active approach (ultrasound) cannot gather information about unrecognizable areas.

Japanese Patent Application Laid-Open No. 2008-175733 JP 2011-220701 A Specification

F. Asano, M. Goto, K. Itou, and H. Asoh, “Real-time sound source localization and separation system and its application on automatic speech recognition,” in Eurospeech 2001, Aalborg, Denmark, 2001, pp. 1013- 1016. CT Ishi, O. Chatot, H. Ishiguro, N. Hagita, “Evaluation of a MUSIC-based real-time sound localization of multiple sound sources in real noisy environments,” in Proc. Of the 2009 IEEE / RSJ Intl. Conf. on Intelligent Robots and System, St. Louis, USA, 2009, pp. 2027-2032.

  However, in such a sound source localization technology, in a real environment, for example, when there is a surface that reflects sound such as a wall, ceiling, glass window, or display around the microphone array, Reflected sound may also be measured. In particular, when the distance between the microphone array and the sound source is large, it is expected that the influence of reflected sound cannot be ignored.

  Therefore, in a real environment, it is desirable to perform sound source localization on the premise that there is an influence of reflected sound.However, a method for estimating the position of a sound source by actively incorporating the influence of reflected sound is , Not necessarily obvious.

  In addition, using reflected sound, it is considered possible to detect other objects approaching from a non-line-of-sight area, but how to apply it to the movement control of a moving body such as a robot. It is not always clear what to do.

  The present invention has been made to solve the above-described problems, and its purpose is to detect the position of another object approaching from a non-line-of-sight area using reflected sound. It is to provide a sound source position estimation device capable of performing the above.

  Still another object of the present invention is to perform movement control using sound source position estimation capable of detecting the position of another object approaching from a region where visibility is not possible using reflected sound. It is to provide a self-supporting mobile that is possible.

  In the present invention, the direction of the sound source is estimated using a plurality of microphone arrays, the direction of the reflected sound is estimated using spatial information, and the information is integrated to perform sound source localization (position estimation in a three-dimensional space). .

  According to one aspect of the present invention, there is provided a sound source position estimation device, which includes a sound sensor array including a plurality of sound sensors, arrangement information of each sound sensor in the sound sensor array, and spatial three-dimensional map information of a measurement environment. A storage device for storing, an incoming sound direction specifying means for executing processing for specifying the direction in which sound arrives at the sound sensor array, based on each of the sound source signals of the plurality of channels from the sound sensor array; A sound source direction estimating means for estimating a plurality of directions toward the sound source in the spatial 3D map information based on the direction of arrival of the generated sound and the spatial 3D map information; For a plurality of directions, sound source position estimating means for estimating the position of the sound source in three dimensions according to the overlap of the extension areas of the route that can be reached by the route due to reflection is provided.

  Preferably, a plurality of sound sensor arrays are provided, and the sound source direction estimating means estimates the direction of the sound source with respect to the reflected sound based on the direction of arrival of the specified sound and the spatial three-dimensional map information, and the sound source position estimation The means estimates the position of the sound source in three dimensions according to the overlap of the extension regions in the direction of the sound source due to the reflected sound from each of the plurality of sound sensor arrays.

  According to another aspect of the present invention, the mobile body is capable of autonomous movement, and includes a moving mechanism, a sound sensor array including a plurality of sound sensors, arrangement information of each sound sensor in the sound sensor array, and a measurement environment. And a storage device for storing spatial three-dimensional map information and a process for specifying a direction in which sound arrives at the sound sensor array based on each of the sound source signals of a plurality of channels from the sound sensor array Sound direction specifying means; sound source direction estimating means for estimating a plurality of directions toward the sound source in the spatial 3D map information based on the direction of arrival of the specified sound and the spatial 3D map information; Sound source position estimating means for estimating the position of the sound source in three dimensions in accordance with the overlap of the extended areas of the path that can be reached by the path due to reflection in a plurality of directions from the sound sensor array toward the sound source; Control means for controlling the movement mechanism, and the control means moves when the position of the sound source estimated by the sound source position estimation means is within the line-of-sight region and within the first predetermined distance. Reduce speed.

  Preferably, when the position of the sound source estimated by the sound source position estimating means is within the line-of-sight region and within a second predetermined distance that is closer than the first predetermined distance, Until it is determined that the collision can be avoided, the vehicle stops and waits.

  According to still another aspect of the present invention, there is provided a method for controlling a moving body capable of autonomous movement, the moving body including a moving mechanism, a sound sensor array including a plurality of sound sensors, and a plurality of channels from the sound sensor array. Based on each sound source signal, an incoming sound direction specifying means for executing processing for specifying the direction in which the sound arrives at the sound sensor array, the specified sound arrival direction, and spatial three-dimensional map information Based on the sound source direction estimation means for estimating a plurality of directions toward the sound source in the spatial three-dimensional map information, and an extension of the path that can be reached by the reflection path for the plurality of directions from the sound sensor array toward the sound source A sound source position estimating means for estimating the position of the sound source in three dimensions according to the overlap of the regions, the step of planning the movement path, and the sound source position estimated by the sound source position estimating means Location is a sight not region, if the first is within the predetermined distance, and a step of reducing the movement speed.

  Preferably, when the position of the sound source estimated by the sound source position estimating means is within a line-of-sight region and within a second predetermined distance that is closer than the first predetermined distance, a collision with the sound source in the movement path Until it is determined that it can be avoided, and the moving object waits.

  According to the present invention, it is possible to estimate a three-dimensional accurate position of a sound source using sounds from a plurality of directions of arrival from the same acoustic source.

  Further, according to the present invention, it is possible to estimate the three-dimensional accurate position of the sound source by using sounds from a plurality of directions of arrival from the same acoustic source, and to control the movement of the self-supporting mobile body. .

It is a conceptual diagram which shows the condition assumed in embodiment. It is a figure which shows the hardware constitutions for radiant sound intensity map preparation among the structures of the robot 1000 of this Embodiment. 2 is a block diagram showing a hardware configuration of a computer system 2000. FIG. It is a figure which shows the external appearance of the robot 1000. FIG. It is a figure for demonstrating the concept of the ranging process by a laser range finder. It is a figure for demonstrating the concept of the specific process of the position of the robot. It is a flowchart for demonstrating the specific process of the position of the robot 1000 using a particle filter. It is a figure for demonstrating the concept of calculation of the state vector sm [t] of a particle. It is a figure which shows the likelihood about each particle based on the distance to a target object. It is a conceptual diagram for demonstrating the procedure which calculates the likelihood of a particle based on the measurement result (ranging data) of a laser range finder. It is a conceptual diagram which shows the procedure of the distance measurement by a laser range finder, and the raycast from a certain particle. It is a figure which shows the search grid about an angle. FIG. 5 is a diagram showing a two-dimensional projection map of a three-dimensional geometric map for one of the test environments and a corresponding octomap. It is a figure which shows the procedure of ray tracing. It is a figure which shows the flow of the movement control of a robot. It is a figure which presents three different trials including three scenarios in a situation where information of reflected sound is used. It is a figure which presents three different trials including three scenarios in a situation where information of reflected sound is not used.

  Hereinafter, the configuration of a sound source position estimation apparatus according to an embodiment of the present invention will be described with reference to the drawings. In the following embodiments, components and processing steps given the same reference numerals are the same or equivalent, and the description thereof will not be repeated unless necessary.

  In the following description, as a sound sensor, a so-called microphone, more specifically an electret condenser microphone will be described as an example, but other sound sensors may be used as long as they can detect sound as an electric signal. Also good.

  In the present embodiment, acquisition of prior knowledge of sound environment and its use are collectively referred to as “sound environment intelligence”. Also, a space-time representation of when, where, and what sound is generated, that is, a map that can identify a sound source that generates sound is called a “sound environment map”. Furthermore, a map including not only the position of such a sound source but also the sound intensity emitted by the sound source is referred to as a “radiated sound intensity map”.

  In a real environment, a plurality of sounds generated at different locations are mixed and observed, so that a simple method of scanning a space with a sound level meter is insufficient for generating a sound environment map. Generating a sound environment map that characterizes the location and type of a sound source that may be useful as prior knowledge of the sound environment requires at least localization and separation of the sound source in addition to spatial information (normal map). It is desirable that the sound source is classified.

  Therefore, in the sound source position estimation apparatus according to the present embodiment, as described below, in order to localize and separate a plurality of sound sources, a microphone array including a plurality of sound sensors is linked to generate sound for a specific sound source in the space. Generate an environmental map to structure the sound environment.

  Here, “sound source localization” means to continuously specify the direction of the sound source, and “sound source position estimation” is based on the direction of the sound source specified by the sound source localization in a predetermined space. It means estimating the position of a three-dimensional sound source.

  In the following, as an embodiment, a sound source position estimation device capable of detecting the position of another object approaching from a region where visibility is not possible using reflected sound will be described.

  In addition, a moving body capable of performing movement control of an autonomous moving body using such sound source position estimation will be described. In the following description, the mobile body will be described using an autonomously movable robot as an example.

  FIG. 1 is a conceptual diagram showing a situation assumed in the present embodiment.

  That is, in the robotic context, as shown in FIG. 1, the acoustic information coming from the occlusion area (non-line-of-sight area) to detect an approaching person and slow down or wait until the person passes. Means to use.

  However, robots usually do not use such information and rely only on sensors (cameras, lasers, ultrasonic sensors) that require the subject to be seen, so they only obtain information from the line of sight .

  Therefore, in the scenario depicted in FIG. 1, the robot may travel to the intersecting location at regular speed and collide with a person.

  The problem is that, for a moving body that can move autonomously, such as a robot wheelchair, and that a person is on board, the person on board may hear that the person is approaching the intersection in the hallway. In particular, it is particularly important from the viewpoint of removing anxiety given to the user and realizing safe movement.

  In other words, in the situation as described above, if the robot does not take any measures against the approaching person, it will be a stressful situation for the rider.

  As will be described below, in the present embodiment, propagation by multipath transmission in the sense that sound is transmitted while reflecting is given information regarding the presence of an approaching person in a line-of-sight region. In other words, the position of the approaching person can be specified using the sound reflected by the environment.

  In order to realize such processing, the robot must be able to detect the reflected sound and evaluate the direction of arrival (DOA).

  Detection of the direction of sound arrival (sound source localization) is performed using a microphone array, using the MUSIC method or other passive acoustic localization methods such as steered response power phase conversion (SRP-PHAT: Steered Response). (Power-Phase Alignment Transform, hereinafter simply referred to as SRP).

  Such a steered response power phase conversion method is disclosed in, for example, the following known documents.

Known Document 1: Japanese Patent Application Laid-Open No. 2012-242597 Known Document 2: MS Brandstein and HF Silverman, “A robust method for speech signal time-delay estimation in reverberant rooms,” in IEEE Conference on Acoustics, Speech, and Signal Processing , ICASSP 1997, 1997, pp. 375-378.
In addition, the robot tracks the acoustic propagation path (going back from the receiving side to the acoustic source side) to find the acoustic source of the incoming reflected sound. Such tracking is called “ray tracing”.

  In the environment, after reflecting one or more times, the sound from the non-line-of-sight area will be processed, so the robot not only has information on the position of the reflector, but also its posture (position and direction). You must know.

  As will be explained later, recent advances in 3D Light Detection and Ranging sensors (LIDAR) have made it possible to create accurate 3D maps in the environment and use them to evaluate reflections. it can.

  The posture of the robot is evaluated by a localization / algorithm based on a particle filter.

  By combining these techniques, the robot can evaluate a set of “sounds arriving on a linear path (called an acoustic ray)” from an acoustic source in the non-line-of-sight region.

  Thereafter, the position of an approaching object (eg, a person) is evaluated by considering the location where these acoustic rays intersect. Here, the “intersection position” does not need to completely intersect, and includes a position where a plurality of acoustic rays are within a predetermined distance range.

  Therefore, the robot can detect not only an approaching person as shown in FIG. 1 but also any object that emits sound.

  Considering sound reflections has to do with multipath transmission of sounds, and in particular early reflections (sounds that reach the listener after a few reflections).

  In order to model the initial reflection, a virtual sound source method is employed that uses ray tracing to calculate the reflection of sound to the surface using a virtual sound source. A virtual acoustic source is obtained by mirroring the acoustic source with respect to the surface. These reflections are often referred to as specular reflections.

  Hereinafter, in this embodiment, when referring to specular reflection, the term “reflection” is simply used.

  In the method of sound source localization according to the present embodiment, an approach using multimodal sensing is used for sound source mapping and safe navigation of an autonomous mobile body such as a robot.

  The mapping of the acoustic source is performed by combining the path measurement information from the robot and the data from the three-dimensional laser scanner. The resulting 3D map is then used for acoustic reflection analysis during robot navigation.

  FIG. 2 is a diagram showing a hardware configuration for creating a radiated sound intensity map among the configurations of the robot 1000 according to the present embodiment. In FIG. 2, the robot 1000 is provided with a moving mechanism 70 (for example, a driving mechanism using two wheels that can operate in a differential manner) for moving the robot, and visually, like a camera. A sensor for detecting the external environment is not shown. In the following example, a human-type autonomous mobile robot driven by two wheels will be described as an example of a “moving body” for creating a radiated sound intensity map. The moving body and the robot having other configurations may be used as long as they can move autonomously in the space without being limited thereto.

  Referring to FIG. 2, robot 1000 has a path length measurement sensor 10 for detecting a moving distance, a speed, an angular velocity, and the like of a moving mechanism 70 that moves the robot by a two-wheel drive mechanism. A laser range finder (LRF) 20 that is mounted on the front and measures the distance to an object that is in front by scanning the laser beam, and mounted on the rear surface of the robot 1000 and scans the laser beam. The rear surface laser range finder (LRF) 30 for measuring the distance to the object existing behind and the 3D light detection distance measuring sensor (hereinafter referred to as 3D) that performs measurement for creating a 3D geometric map. LRF) 32, signals from the path length measurement sensor 10, the front surface LRF 20, and the rear surface LRF 30, and the three-dimensional LRF 3 The signals from, and output as data through the bus 1060, or, and a sensor output board 40 for transmitting a command from the bus 1060 to the sensors.

  The robot 1000 further functions as a working memory and includes a memory 1050 including a RAM (Random Access Memory), a program (not shown) for operating the robot 1000, a three-dimensional geometric map 1102, and a three-dimensional Non-volatile storage for storing the octree data 1106 created from the geometric map, the surface normal lookup table 1108 for calculating and storing the surface normal of the object in space, the acoustic source map 1104, and the like. Device 1100. As the nonvolatile storage device 1100, a hard disk may be used as long as it is a randomly accessible storage device, or an SSD (Solid State Drive) or the like may be used. The “three-dimensional geometric map” is data that geometrically represents the position of a wall or a fixed / semi-fixed object in the three-dimensional space where the robot 1000 moves on a map. Say. The octree data 1106 is a kind of spatial index used for optimizing a spatial query in a spatial database (a database that stores information such as the position of objects in the space), and is a kind of tree structure. The node has a maximum of 8 child nodes. This data is used when recursively dividing a three-dimensional space into eight octants. In addition, as described above, the surface normal lookup table 1108 calculates the normal direction in advance for each point on the surface of a wall or a fixed / semi-fixed object in a three-dimensional space. This table is used when calculating the reflection from the object, as will be described later.

  The robot 1000 further includes a microphone array 1052 for measuring the direction of the sound source (sound arrival direction) and the acoustic power from the direction, and a signal for transmitting the signal from the microphone array 1052 to a signal transmitted to the bus 1060. A voice input / output board 1054 and a processor 1010 which is an arithmetic unit for controlling the operation of the robot 1000 and executing creation of a three-dimensional geometric map and creation and update of an acoustic map are provided. Although not shown in FIG. 14, the audio input / output board is also connected to a speaker that reproduces the audio generated by the processor 1010 for communication with the user.

  The processor 1010 creates a three-dimensional geometric map based on the signals from the path measurement sensor 10, the front surface LRF 20 and the rear surface LRF 30, and the signals from the three-dimensional LRF 32, based on a program stored in the nonvolatile storage device 1100. A geometric map creation unit 1030 to perform, a steered response power acquisition unit 1050 for executing processing for acquiring a spectrum of a sound source power based on a signal from the microphone array 1052, and a sound source direction (sound of the sound) based on the sound source power spectrum. In order to specify the posture (position / direction) of the robot on the basis of signals from the arrival sound direction distribution estimation unit 1060 for estimating the arrival direction on the robot coordinates, and the path length measurement sensor 10, the front surface LRF 20, and the rear surface LRF 30. Mobile body posture identification processing unit 1040 Based on the estimation results of the bot position and three-dimensional geometrical map 1102 arrival sound orientation distribution estimation unit 1060, and a sound source position estimation processing unit 1070 for estimating the sound source position.

Here, in the case of sound source direction estimation using the phase difference, it is possible to perform sound source direction estimation using the SRP-PHAT method (Steered Response Power-Phase Alignment Transform).
[Realization by computer]
The functions of the processor 1010 and the nonvolatile storage device 1100 described above are actually realized by the cooperation of hardware and software by computer hardware and a computer program executed by the computer hardware. The operation of the computer program for realizing the sound source localization processing function will be briefly described below.

  FIG. 3 is a block diagram showing the hardware configuration of a computer system 2000 for executing such a computer program.

  As shown in FIG. 3, in addition to the disk drive 2030 and the memory drive 2020, the computer main body 2010 constituting the computer system 2000 includes a CPU (Central Processing Unit) 2040 and a ROM (Read Only memory) 2060 and RAM (Random Access Memory) 2070, a non-volatile rewritable storage device such as a hard disk (or SSD) 2080, and a communication interface 2090 for performing communication via a network , And microphone arrays MC1 and MC2, and an audio input interface 2092 for exchanging signals. The disc 2030 is loaded with an optical disc such as a CD-ROM 2200. A memory card 2210 is attached to the memory drive 2020.

  As will be described later, when the program of the apparatus that performs the source position estimation process operates, the database that stores information that is the basis of the operation is described as being stored in the hard disk 2080.

  In FIG. 3, the CD-ROM 2200 is assumed as a medium capable of recording information such as a program installed in the computer main body. However, other media such as a DVD-ROM (Digital Versatile Disc) is used. Alternatively, a memory card or a USB memory may be used. In that case, the computer main body 2010 is provided with a drive device capable of reading these media.

  The input devices 2100 and 2110 and the display display 2120 are provided as necessary.

  The main part of the apparatus that performs the sound source position estimation process is configured by computer hardware and software executed by the CPU 2040. Generally, such software is stored and distributed in a storage medium such as a CD-ROM 2200, read from the storage medium by a disk drive 2030 or the like, and temporarily stored in the hard disk 2080. Alternatively, when the device is connected to the network 310, it is temporarily copied from the server on the network to the hard disk 2080. Then, the data is further read from the hard disk 2080 to the RAM 2070 in the memory and executed by the CPU 2040. In the case of network connection, the program may be directly loaded into the RAM and executed without being stored in the hard disk 2080.

  A program for functioning as a device for performing sound source position estimation processing does not necessarily include an operating system (OS) that causes the computer main body 2010 to execute functions such as an information processing device. The program only needs to include an instruction portion that calls an appropriate function (module) in a controlled manner and obtains a desired result. How the computer system 20 operates is well known and will not be described in detail.

  Further, the computer that executes the program may be singular or plural. That is, centralized processing may be performed, or distributed processing may be performed.

  Further, the CPU 2040 may be a single processor or a plurality of processors. That is, it may be a single core processor or a multi-core processor.

  Note that a database that stores information that is the basis of the operation of the program of the device that performs sound source position estimation processing may be stored in an external storage device connected via the interface 2090. For example, when connected to an external server via a network, a database that stores information serving as a basis of operation may be stored in a storage device such as a hard disk (not shown) in the external server. In this case, the computer 2000 operates as a client machine, and exchanges such database data with an external server via a network.

  FIG. 4 is a diagram illustrating an appearance of the robot 1000.

  Referring to FIGS. 2 and 4, a front laser range finder (LRF) 20, a rear laser range finder (LRF) 30, and a three-dimensional LRF 32 are provided on the front and rear surfaces of the robot 1000. . Further, a path length measurement sensor 10 is attached to a wheel that is the moving mechanism 70.

  The intensity of the reflected sound from the object (obstacle) was evaluated using the sound intensity at the platform position (evaluated by the maneuvered response power algorithm) and the object (laser range finder) ) By combining the distance to.

  By combining the platform attitude obtained from the localization algorithm based on particle filter and the sound intensity evaluated at the platform position, the position of the sound source of the reflected sound from the object covers the environment Identified in a grid map cell.

  In the present embodiment, various information included in the sound itself is used to detect a moving object outside the field of view of the robot at the intersection. The assumption is that these mobiles are emitting sound and these sounds are reflected back to the environment.

  Based on the acoustic propagation characteristics, the reflected sound reaches the microphone array 1052 mounted on the robot.

  Therefore, the acoustic source position estimation processing unit 1070 mounted on the robot evaluates the arrival direction (DOA) of the reflected sound that arrives at a place where passages such as a corridor intersect.

  In order to know the arrival direction of the incoming sound and the posture of the robot and find the position in the environment where the received sound is emitted, the ray tracing processing unit 1072 performs ray tracing.

  If the properties of the sound reflector, especially the normal of the reflecting surface, can be determined, it is possible to calculate the direction of arrival of the reflected sound (at the reflection point). By repeating the ray tracing and reflection calculation procedure, the acoustic ray can be determined from the robot position to the acoustic source.

However, with one acoustic ray, it is not possible to determine the number of reflections and the exact location of the acoustic source. For this reason, the location of the acoustic source is evaluated by calculating a plurality of acoustic rays and determining their intersection. This requires that the audio scan exhibit several maxima corresponding to different reflections of the same acoustic source.
(Create a three-dimensional geometric map of the environment)
As a prerequisite for generating a radiated sound intensity map, a three-dimensional geometric map is prepared in advance that describes the environment and allows the mobile platform to locate itself. is necessary. Creation of such a three-dimensional geometric map is executed by the geometric map creation unit 1030.

  Such a geometric map represents the environment by an “occupied grid”. That is, in such a map, each cell constituting the grid is classified into the following three types.

  i) The cell is empty (open space).

  ii) The cell is occupied (there are walls and structures).

  iii) The cell has not been investigated (it is unknown whether it is empty or occupied).

  The geometric map is pre-built using a 3D toolkit library framework.

  Here, such a framework is disclosed in the following documents.

Known Document 3: D. Borrmann, J. Elseberg, K. Lingemann, Andreas Nuchter, and J. Hertzberg, “The Efficient Extension of Globally Consistent Scan Matching to 6 DoF,” in Proceedings of the 4th International Symposium on 3D Data Processing, Visualization and Transmission (3DPVT '08), Atlanta, USA, June 2008, pp. 29-36.
Therefore, to explain briefly, first, the path measurement and LRF data necessary to construct a geometric map are transferred to the mobile platform by the remote control device operated by the user. Collected by passing.

  Subsequently, mapping technology based on simultaneous positioning and mapping (SLAM) for creating geometric maps based on the information collected in this way has been well studied and completed. It has been applied in practice.

  In this embodiment, the technique is used to generate a grid map.

  For example, using an iterative closest point (ICP) based SLAM, a 3D Toolkit library framework is used to modify the robot trajectory and align the laser rangefinder scan with the map.

  With the resulting aligned scan, an occupancy grid map is created.

  Such an occupation grid map is disclosed in, for example, the following documents.

Known Document 4: A. Elfes, “Using occupancy grids for mobile robot perception and navigation,” Computer, vol. 22, no. 6, pp. 46-57, June 1989.
Therefore, in the present embodiment, “matching the scan of the laser range finder to the map” means matching the direction and position of the scan of the laser range finder and the positioning position in the design drawing. The “occupied grid” is a fixed object (a fixed object such as a wall or a semi-fixed object such as a television set) on a geometric map, and is the time to be measured. A grid that is occupied by what is considered fixed relative to the spacing.

  As will be described later, in the present embodiment, the position of the acoustic source in the space is specified by combining the direction of the acoustic source and information related to the “occupied grid”.

  The resolution of the map cell was set to 5 cm × 5 cm × 5 cm.

  After the 3D geometric map was obtained, the 3D geometric map was converted to an octree using an octomap framework.

  Such an octomap framework is disclosed in the following document, for example.

Known Document 5: A. Hornung, KM Wurm, M. Bennewitz, C. Stachniss, and W. Burgard, “OctoMap: An efficient probabilistic 3D mapping framework based on octrees,” Autonomous Robots, 2013, software available at http: // octomap.github.com. [Online]. Available: http: //octomap.github.com
(Identification of robot position)
The purpose of specifying the position of the robot 1000 is to accurately evaluate the posture (location and direction) of the robot 1000 in the geometric map representing the environment.

  The moving body posture identification processing unit 1040 is performed by a position identification algorithm based on a particle filter that combines the path length measurement (wheel encoder) and information from the LRFs 20 and 30.

  One set of M particles approximates the probability density function of the posture of the robot 1000.

  The likelihood of particles during the estimation step is based on a ray casting approach.

  The posture {x (k), y (k), θ (k)} of the robot 1000 is given from a weighted average of particle postures (or the most promising particle posture may be used).

  FIG. 5 is a diagram for explaining the concept of distance measurement processing by the laser range finder.

  The laser range finder is a device that irradiates laser light from the robot 1000 at an angle b and measures the distance z to the target object at the angle b. By scanning while changing the angle b, the distance to the object within a certain angle range can be acquired.

  As shown in FIG. 5, it is possible to acquire the distance to the object existing at 360 degrees around the robot by the front surface LRF 20 and the rear surface LRF 30 in each one scan.

  FIG. 6 is a diagram for explaining the concept of the process for specifying the position of the robot 1000.

  As shown in FIG. 6A, when only the path sensor 10 is used to specify the position of the robot 1000, the robot 1000 actually moves even if the start point and the end point coincide with each other. The position of the robot specified inside will shift.

  On the other hand, as shown in FIG. 6B described below, when the distance sensor 10 and the distance measurement data of the front surface LRF 20 and the rear surface LRF 30 are combined, even at a position specified inside the robot, the starting point And the end point match.

  Thus, in order to combine the distance sensor 10 and the distance measurement data of the front surface LRF 20 and the rear surface LRF 30, as described below, in this embodiment, the weighted M pieces are used in evaluating the robot position. Use a particle filter approach with multiple particles.

  Each of the m particles includes, as its attributes, the candidate position {xm (k), ym (k)} and candidate direction {qm (k)} of the robot 1000 and the particle weight {wm (k)}. State vectors sm [t] = {xm (k), ym (k), qm (k), wm (k)} are associated.

  While the robot 1000 moves, each particle also moves based on path length measurement and a stochastic motion model (which describes uncertainty in the robot motion).

  Further, in the process of calculating the weight for the particle, the particle filter estimates the posterior probability in consideration of the likelihood based on the measurement by the laser range finder.

  FIG. 7 is a flowchart for explaining the process of specifying the position of the robot 1000 using the particle filter.

  Referring to FIG. 7, when the process is started, first, the moving body posture identification processing unit 1040 initializes the attribute of each particle (S100).

  Subsequently, the moving body posture identification processing unit 1040 increments the variable m sequentially from 1 to the maximum value mmax of m, and obtains the particle state vector sm [t], the path measurement result, the motion model, and the previous time. Calculation is performed based on the state vector sm [t-1] (loop of S102, S104, S106, and S108).

  FIG. 8 is a diagram for explaining the concept of calculating the particle state vector sm [t].

  As shown in FIG. 8, the path sensor 10 measures the speed v_r of the right wheel and the speed v_l of the left wheel, so that the moving body posture identification processing unit 1040 causes the speed V of the robot 1000 to be measured. And each speed R can be calculated. As a result, the state vector sm [t] of the particle after the lapse of Δt time is calculated from the state vector sm [t−1] at the previous time. Each particle is different not only in position but also in direction.

  Returning to FIG. 7, subsequently, the robot 1000 is controlled by the processor 1010 to initialize the variable m for counting particles to 1 (S108), and initialize the measurement angle and the variable k (S110). ) After that, distance measurement to the object is performed by the laser range finders 20 and 30 every predetermined angle Δθ from the angle θ0 (S112).

  Subsequently, the moving body posture identification processing unit 1040 calculates the likelihood for each particle based on the distance to the target object at the time t in each direction at time t (S114).

  FIG. 9 is a diagram showing the likelihood of each particle based on the distance to such an object.

  The likelihood based on the distance measurement of such a laser range finder is not intended for particles, but is described in the following documents.

Known Document 6: S. Thrun, W. Burgard, and D. Fox, Probabilistic Robotics (Intelligent Robotics and Autonomous Agents), The MIT Press, 2005.
As shown in FIG. 9A, when the distance to the object is zt * by the laser range finder, such likelihood (posterior probability distribution) is obtained by superimposing the following probability distributions.

i) Gaussian distribution centered on the measurement distance zt * in consideration of measurement noise ii) Probability that an unexpected object is detected when the robot moves, and a distribution that decreases exponentially with distance iii ) Probability that the distance measurement data will be the maximum when there is no object within the maximum distance measurement range iv) Probability distribution corresponding to white noise due to random factors However, in the following description, for the sake of simplicity The degree will be described assuming that the degree is expressed as shown in FIG.

  FIG. 10 is a conceptual diagram for explaining a procedure for calculating the likelihood of particles based on the measurement result (ranging data) of the laser range finder.

  Consider the case where the distance from the actual position of the robot 1000 to the object at the angle θk is measured as z * from the actual position of the robot 1000 as shown in FIG.

  At this time, with respect to the particle 1, it is calculated that the distance to the object at the angle θk is Z1 * in the geometric map, assuming that a light ray is emitted from its position and orientation. Assume that the distance to the object at the angle θk is calculated to be Z2 * in the geometric map, assuming that a ray is emitted from the direction.

  Here, the process of irradiating light rays virtually from the particles to the object also corresponds to the fact that irradiating light rays at a predetermined angle in the actual scanning operation of laser light is called “ray casting”. This is called “Raycast”.

  At this time, based on the probability distribution shown in FIG. 9, the moving body posture identification processing unit 1040 calculates the likelihood of the particle 1 as L1, as shown in FIG. 10, and sets the likelihood of the particle 2 as L2. Is calculated (S114).

  If the angle θk does not reach the maximum value (S116), the moving body posture specifying processing unit 1040 updates the angle θk by Δθ (S118), and measures the distance to the object at the angle θk with the laser range finder. Based on the calculation of the distance to the object at the angle θk for each particle, the likelihood is calculated and integrated by the probability distribution shown in FIG. 9 (S114).

  FIG. 11 is a conceptual diagram showing a procedure of distance measurement by a laser range finder and ray casting from a certain particle.

  As shown in FIG. 11, the result of ranging from the robot 1000 using the laser range finder while changing the angle, and the distance to the target calculated on the geometric map by ray casting from each particle, Thus, the likelihood of particles at each angle is calculated, and the likelihoods obtained for the angles are integrated.

  Returning to FIG. 7 again, by repeating such processing until the angle θk reaches the maximum value, the moving body posture identification processing unit 1040 determines the weight of each particle in proportion to the height of likelihood ( S120).

  The processes from step S110 to step S120 as described above are repeated until the maximum number mmax of particles is reached while incrementing the value of the variable m (loop from S110 to S124).

  Further, the moving body posture identification processing unit 1040 estimates the probability density distribution of the robot position from the current state represented by the particle and its weight (S126). Furthermore, the moving body posture identification processing unit 1040 estimates the current state (position and orientation) of the robot using the estimated probability density distribution (S128).

  Then, the moving body posture identification processing unit 1040 performs resampling processing on the particle filter using the estimated probability density distribution (S130). That is, for example, particles are restored and extracted with a probability proportional to the probability density distribution. That is, the extraction frequency of particles with a large probability is increased, and particles with a small probability are extinguished to obtain a particle set at time t.

  Thereafter, the process returns to step S102.

  As a result of the above processing, after the ray casting, the particles more accurately reflecting the state of the robot have a higher likelihood score and are assigned more weight.

(Intensity of sound in the receiving direction)
The direction of arrival of sound incident on the microphone array 1052 is evaluated by the SRP algorithm.

  The steered response power acquisition unit 1050 evaluates the sound intensity in a certain direction as a function of the arrival angle using a steered response power (SRP) algorithm, as described in the known document 2 and the like.

  As described below, in the present embodiment, the SRP method is executed based on a delay and a sum beamformer.

  After applying a short-time Fourier transform (STFT) to the observed signal sampled at 48 kHz (analysis window is 25 milliseconds in length and the window shift is 10 milliseconds), the processing is performed in the frequency domain. Done. Here, “short-time Fourier transform (STFT)” refers to multiplying a function while shifting a window function and performing Fourier transform on the function. Theoretically, in order to obtain a Fourier coefficient, integration must be performed over an infinite interval, but in order to obtain a Fourier coefficient from an experimental value or the like, the range must be divided. Multiplication is performed by eliminating a discontinuous element as much as possible by multiplying by a function whose value is 1 near the center and converges to 0 outside the range. This is a short-time Fourier transform. At this time, the function to be multiplied is called a window function.

After the STFT, at the sampling time t, the observed signal from the Q microphones at the frequency f is displayed as U ₁ (f, t),..., U _Q (f, t).

  Phase transformation is applied to the frequency domain signal before beamforming by performing the following equation (1).

This is because if phase transformation is generally performed in SRP, it becomes less sensitive to reverberant sound.

  To accurately evaluate the sound from a spatial location (represented in the reference frame of the array described by polar coordinates {ρ, θ, φ}), since the geometry of the microphone array is known precisely It is possible to steer the array using spatial beamforming.

  The beamforming output of frequency f is expressed by the following equation (2).

Here, the following expression represents the delay in the q-th microphone for the location {ρ, θ, φ}.

This is because if phase transformation is generally performed in SRP, it becomes less sensitive to reverberant sound.

  To accurately evaluate the sound from a spatial location (represented in the reference frame of the array described by polar coordinates {ρ, θ, φ}), since the geometry of the microphone array is known precisely It is possible to steer the array using spatial beamforming.

  The SRP algorithm evaluates the beamformer output power for a set of N locations.

  Here, it is assumed that the far field condition is satisfied (ρ is larger than the aperture of the array). Therefore, a set of N positions is defined by angles {θn, φn}, nε [0, 2π], where nε [1, N] (steps Δθ, Δφ).

  FIG. 12 is a diagram showing a search grid for such an angle.

Referring to FIG. 12, the search grid has an equilateral triangle shape (the angle between two vertices is approximately 3 °).
The sound power is then calculated within the [1000, 5000] Hz frequency band corresponding to the discrete frequencies f ₁₀₀₀ and f ₅₀₀₀ by the following equation:

Here, #F is the number of discrete frequencies in the band [1000, 5000] Hz.

  Then, the time average represented by the following equation is applied to combine L (= 10 STFT) frames.

Here, k is an index corresponding to the block of the L frame.

  That is, the k-th audio scan is a set of N angles {θn, φn} and sound power Jn (k) associated with them.

  The frequency of the audio scan is 10 Hz (L (= 10 STFT) window shifted by 10 milliseconds).

  The direction of arrival of the signal incident on the microphone array can be obtained by searching for the maximum value during the audio scan.

For example, in FIG. 12, the audio scan clearly shows two maxima.
(Ray tracing in 3D geometric maps and reflection calculations)
As described above, in this embodiment, beam forming is performed using the SRP algorithm.

  Since the direction of interest is in front of the robot, audio scanning does not lose its generality when constrained to the front of the robot. The audio scan is limited to a range of the yaw angle θn and the pitch angle φn that satisfies such a condition.

  An audio scan is considered with robot center coordinates (hereinafter referred to as robot coordinates) in the range between the yaw angles θmin and θmax and the pitch angles φmin and φmax.

  As described above, the SRP method gives a distribution of audio power in the microphone array coordinates.

  This audio power data is converted into robot coordinates by using the attitude of the microphone array relative to the robot.

  On the other hand, the accurate posture (position and direction) of the robot in the three-dimensional geometric map is obtained by the position specifying algorithm based on the particle filter as described above.

  By combining the robot attitude and the results of the SRP process, an audio power distribution in global coordinates around the robot can be obtained.

  For ray tracing, first, local maximum values above a certain threshold are considered. Each yaw angle and pitch angle of the local maximum is obtained.

  A microphone array 1052 in the global coordinate system is obtained by conversion from the robot coordinate system to the global coordinate system, and ray casting is performed in the three-dimensional geometric map from the exact position of the microphone array 1052. FIG. 12 shows the appearance of such a yaw angle and pitch angle and the maximum value during the trial, and the state of ray tracing by ray casting toward the maximum value.

  FIG. 13 shows a map obtained by projecting a three-dimensional geometric map for one of the test environments, two-dimensionally, and a corresponding octomap created by the geometric map creation unit 1030 as described above. The upper diagram in FIG. 13 shows a two-dimensional map, and the lower diagram shows an octomap.

  Ray tracing by ray casting as described above is executed in a three-dimensional map represented in an octree format. As described above, the octomap framework is used for operations and operations on octree maps.

  FIG. 14 is a diagram showing a ray tracing procedure.

  In FIG. 14, the processing performed in three dimensions is actually projected in two dimensions.

  As shown in FIG. 14, with respect to the coordinates and the direction vector (d) of the starting point of ray casting in the three-dimensional geometric map, ray casting is performed from the starting point in the direction of the vector d until it hits the occupied cell in the occupation map. Executed. A cell that is hit by a raycast is considered an obstacle.

  Once an obstacle that is the output of ray casting is obtained, reflections from that obstacle must be considered in order to perform further ray tracing.

  The normal direction of the obstacle surface is obtained from a predetermined look-up table 1108 created by surface normal estimation in off-line processing. For every point in the three-dimensional geometric map, the normal direction is pre-calculated to aid the real-time performance of the system.

  The normal vector at the point where the reflection occurs uses a point cloud library that uses principal component analysis (PCA) of the covariance matrix created from the nearest neighbors of the point whose surface normal is pre-evaluated And evaluated.

  Such a procedure is disclosed in the following document, for example.

Known Document 7: RB Rusu and S. Cousins, “3D is here: Point Cloud Library (PCL),” in IEEE International Conference on Robotics and Automation (ICRA), Shanghai, China, May 9-13 2011.
The direction vector (r) of the reflection line is obtained from the incident line vector (i) and the surface normal vector (n) transmitted from the starting point to the obstacle by the following equation.

Now that obstacle becomes the new starting point for the next level of ray casting. The direction vector of ray casting is defined by the reflection angle of the reflection vector r.

  Again, the process of ray casting is repeated, identifying sequentially where new acoustic rays from the obstacle hit the occupied cells in the occupation map. This process is repeated until a predetermined number of reflections from the object are reached.

  In the following description, it is assumed that ray tracing is executed up to four reflections in consideration of error accumulation due to reflection.

  This ray tracing in the 3D geometric map gives an estimate of what path the sound from the acoustic source has taken before reaching the microphone array 1052.

  For example, if the sound is generated from only one moving acoustic source, the acoustic rays from the robot will ideally all source the sound received by the microphone array 1052 on the robot. They will gather at one common point that intersects at the point they represent.

  It is traced from different azimuth orientations (representing different reflections of sound from an acoustic source) to a predetermined number of reflections.

  Assume that each audio scan gives X maxima having a value Jn (k) greater than the threshold p. Each acoustic ray is traced in the direction indicated by these X maxima.

  Consider X acoustic rays as a pair, and calculate the point of intersection of each pair. Since the sound received by the microphone array 1052 is generated, for example, from a contact point between a person's foot and the ground, ideally, X acoustic rays have one common point that intersects at this point. Will have.

  In practice, there may not be a proper intersection of these acoustic rays since they are projected in a 3D map and the reflective surface is irregular. However, these acoustic rays should pass close to each other. Therefore, when an acoustic ray is present in proximity within a predetermined range, processing is performed that regards it as an intersection.

  Then, considering the points closest to the intersecting points of the acoustic ray, these points are plotted in a geometric map. A plot of these intersecting points sequentially over time gives the direction and speed of movement of the acoustic source (this process is called “reflection analysis”).

Using this information, the robot will be able to understand the situation and take communication actions for safe navigation.
(Identification of acoustic source position and robot movement control)
(Experimental settings)
In order to confirm the above-described approach in the experiment, an intersection in the corridor that cannot see the passage where the LRF mounted on the robot intersects was selected.

  Two different types of corridor intersections, a “T” shape and another “+” shape intersection, were considered. The walls of the intersection are all made of concrete and the surface is almost flat. The ceiling is a porcelain material.

  The ambient noise at the experimental site is approximately 42 dB due to the centralized air conditioning system.

  Since the intent of the experiment is to detect human footsteps and sounds from cart wheels, the search range for the yaw angle is between θmin = −75 ° and θmax = 75 °. The pitch angle was between φmin = −45 ° and φmax = 15 °.

  The threshold for the local maximum was set to 28, which was determined experimentally for this system.

  FIG. 15 is a diagram illustrating a flow of robot movement control.

  Referring to FIG. 15, first, the movement control unit 1090 of the robot 1000 plans a movement path from the start position to the target position using a known wide area planner algorithm (S300).

  Subsequently, the movement control unit 1090 sets the moving speed of the robot 1000 to a first predetermined speed V1, for example, 0.5 m / s (S302), and at the speed of the first predetermined speed V1, It moves along the planned route (S304).

  When the robot 1000 is within a predetermined distance L1 from the intersection location in the corridor (for example, approximately 4-5 meters from the intersection location), the movement control unit 1090 receives the information from the acoustic source position estimation processing unit 1070. Based on the estimation result, voice information is used to identify an obstacle out of the line-of-sight area (S310).

  When the sound power of the acoustic source, for example, the output of SRP is higher than a threshold value of a predetermined value (here, 28) (S314), the movement control unit 1090 sets the moving speed from the first predetermined speed. Is set to a lower second predetermined speed, for example, 0.2 m / s (S316). If the SRP output is lower than the predetermined threshold value, or after setting the moving speed to the second predetermined speed, the process returns to step S304.

The movement control unit 1090 uses the voice information when the robot 1000 is within a predetermined distance L1 from the intersection in the hallway, and further, within a predetermined distance L2 (<L1) from the intersection in the hallway (for example, 2 If it is within 5 m (S312), if the sound power of the acoustic source, for example, the output of SRP is higher than a threshold value of a predetermined value (here, 28) (S320), it is moved by reflection analysis The moving direction and speed of the incoming object are calculated (S322). When the possibility that the robot 1000 and the approaching object collide with each other on the movement path is equal to or higher than a predetermined level (S324), the movement control unit 1090 is safe for the robot to move. The robot 1000 is stopped and waited until the position is reached (S326). If the output of SRP is lower than a predetermined threshold value, or if the possibility of collision with an approaching object is less than a predetermined level, the process returns to step S302.
(Experimental result)
The approach in the present embodiment performed support for the robot to safely move to the destination while avoiding any collision during all trials.

  The experiment was conducted with walking people and people pushing indoor carts. The robot showed intelligent behavior by reducing the speed when the reflected sound from the fork in the hallway was heard. When the moving body is identified in the direction of the destination, the robot stops and waits until the moving body passes the intersection.

  Figures 16 and 17 present six different trials involving three scenarios.

  Robot trajectories are shown with shades of speed, and important speed changes are shown with arrows in the figure.

  The first scenario was a situation where there were no mobiles coming to the hallway intersection. The second scenario was a person walking in the hallway from left to right in robot coordinates. The third scenario was a person pushing an indoor cart and moving from right to left.

  FIG. 16 shows a case of robot movement control using acoustic reflection information, and FIG. 17 shows a case of robot movement control without using acoustic reflection information.

  FIGS. 16 (a) and 17 (d) show a situation where there is no moving body coming to the intersection of the hallway, and FIGS. 16 (b) and 17 (e) show that the person is left in the hallway in the hallway. It is a situation to walk to the right. FIG. 16C and FIG. 17F show a situation where a person pushes the indoor cart and the person moves from right to left.

  The X axis and the Y axis indicate global coordinates. The start position is (3500,300) and the destination is (3200,0).

  In the case of not using the information of the acoustic reflection, a collision between the robot and the moving object (person) occurs, and the location of the collision is indicated by X in the figure.

Apart from the results shown, the experiment was repeated for different combinations of direction of movement and whether it was a walking person or a cart, and in all cases the robot was as expected It was confirmed that the operation was performed to avoid collision.
Furthermore, the results were almost similar in both the “T” -shaped intersection and the “+”-shaped intersection in the hallway.

  That is, by predicting a moving body that may suddenly appear at an intersection in a corridor using a reflected sound, the robot can avoid a collision and reach a target position.

  Note that the sound source position estimation apparatus and the sound source position estimation method of the present embodiment as described above have the same functions as the sound source position estimation processing unit 1070 realized by the processor 1010, as in the first embodiment. Thus, it can be realized by processing by a computer.

  As described above, if the sound source position estimation device, the sound source position estimation method, and the sound source position estimation program according to the present embodiment are used, the position of another object approaching from a non-line-of-sight region using the reflected sound Can be detected.

  In addition, in the moving body of the present embodiment, a safe moving body can be used by using a sound source position estimation that can detect the position of another object approaching from a non-line-of-sight region using reflected sound. The control method can be realized.

  The embodiment disclosed this time should be considered as illustrative in all points and not restrictive. The scope of the present invention is defined by the terms of the claims, rather than the description above, and is intended to include any modifications within the scope and meaning equivalent to the terms of the claims.

  10 path length measurement sensor, 20 front laser range finder, 30 rear laser range finder, 32 three-dimensional laser range finder, 40 sensor input / output board, 1050 memory, 1060 bus, 1000 ′ robot, 1010 processor, 1030 geometric map creation unit, 1040 Mobile body posture identification processing unit, 1050 steered response power acquisition unit, 1052 microphone array, 1054 voice input / output board, 1060 arrival sound direction distribution estimation unit, 1070 sound source position estimation processing unit, 1072 ray tracing processing unit, 1100 Non-volatile memory, 1102 3D geometric map, 1104 acoustic source map, 1106 octree data, 1108 surface normal lookup table.

Claims (6)

A sound sensor array including a plurality of sound sensors;
A storage device for storing information on the arrangement of each sound sensor in the sound sensor array and spatial three-dimensional map information of the measurement environment;
Based on each of the sound source signals of a plurality of channels from the sound sensor array, arrival sound direction specifying means for executing processing for specifying the direction in which sound arrives at the sound sensor array;
Sound source direction estimating means for estimating a plurality of directions toward the sound source in the spatial three-dimensional map information based on the identified direction of arrival of the sound and the spatial three-dimensional map information;
Sound source position estimating means for estimating the position of the sound source in three dimensions according to the overlap of extension areas of a path that can be reached by a path due to reflection in a plurality of directions from the sound sensor array toward the sound source; A sound source position estimating apparatus. A plurality of the sound sensor arrays are provided,
The sound source direction estimating means estimates the direction of the sound source with respect to the reflected sound based on the identified direction of arrival of the sound and the spatial three-dimensional map information,
2. The sound source according to claim 1, wherein the sound source position estimating unit estimates the position of the sound source in three dimensions in accordance with an overlap of extension regions in the direction of the sound source due to the reflected sound from each of the plurality of sound sensor arrays. Position estimation device. A mobile object capable of autonomous movement,
A moving mechanism;
A sound sensor array including a plurality of sound sensors;
A storage device for storing information on the arrangement of each sound sensor in the sound sensor array and spatial three-dimensional map information of the measurement environment;
Based on each of the sound source signals of a plurality of channels from the sound sensor array, arrival sound direction specifying means for executing processing for specifying the direction in which sound arrives at the sound sensor array;
Sound source direction estimating means for estimating a plurality of directions toward the sound source in the spatial three-dimensional map information based on the identified direction of arrival of the sound and the spatial three-dimensional map information;
Sound source position estimating means for estimating the position of the sound source in three dimensions according to the overlap of the extended areas of the path that can be reached by the path due to reflection, for a plurality of directions from the sound sensor array toward the sound source;
Control means for controlling the moving mechanism,
The control means reduces the moving speed when the position of the sound source estimated by the sound source position estimating means is within a line-of-sight region and within a first predetermined distance.   When the position of the sound source estimated by the sound source position estimating means is within a line-of-sight region and within a second predetermined distance that is closer than the first predetermined distance, the control means The moving body according to claim 3, which stops and waits for the moving body until it is determined that a collision can be avoided. A method of controlling a mobile body capable of autonomous movement, wherein the mobile body is based on each of a moving mechanism, a sound sensor array including a plurality of sound sensors, and a plurality of sound source signals from the sound sensor array. Based on the incoming sound direction specifying means for executing processing for specifying the direction in which sound arrives at the sound sensor array, and the specified direction of arrival of the sound and spatial three-dimensional map information, the space 3 Sound source direction estimation means for estimating a plurality of directions toward the sound source in the three-dimensional map information, and an overlap of extension areas of the route that can be reached by the path due to reflection for the plurality of directions from the sound sensor array toward the sound source And a sound source position estimating means for estimating the position of the sound source in three dimensions,
Planning a route of travel;
And a step of reducing the moving speed when the position of the sound source estimated by the sound source position estimating means is within the line-of-sight region and within the first predetermined distance.   When the position of the sound source estimated by the sound source position estimating means is within a line-of-sight region and within a second predetermined distance that is closer than the first predetermined distance, The method of controlling a moving body according to claim 5, wherein the moving body is stopped and waited until it is determined that a collision can be avoided.