JPH0772924A - Travel controller for unmanned travelling object - Google Patents

Abstract

PURPOSE:To provide a travel controller for an unmanned travelling object capable of safely and surely running in a fluidly changing and dangerous site where is a natural disaster-stricken area, etc., by unmanned control, and easily assembling a sensor expected to be developed newly in the future even when it is in the middle of the development. CONSTITUTION:Data representing a position on the map of a prescribed route is inputted, and a travel command including a direction in which the unmanned travelling object 1 should run is inferred by a perspective inference part 15 based on the position data of the prescribed route and the present location of the unmanned travelling object 1. While, an obstruction that exists in the periphery of the unmanned travelling object 1 is detected with different detecting ranges by the sensors 5, 6, and the travel command is inferred by a local inference part 16 and a proximity inference part 17 based on detected results, respectively. The travel command is decided by an action judging part 19 based on those inferred results, and the travel of the unmanned travelling object 1 is controlled based on a decided travel command.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a travel control device for an unmanned vehicle, and more particularly to a travel control device capable of safely and reliably performing work in a dangerous natural disaster area.

[0002]

2. Description of the Related Art Japanese Patent Application No. 3-121296 discloses an invention relating to a tracked vehicle having a four-wheel truck frame structure suitable for disaster relief. According to this invention, it is possible to traverse a natural disaster area in Japan by the moving mechanism of the tracked vehicle according to the invention. However, although the tracked vehicle according to the present invention can be remotely controlled, it is a semi-autonomous vehicle that is accompanied by a rescue team and runs through a natural disaster area. In the field of indoor automated guided vehicles, various types of sensors that detect obstacles and the like around the guided vehicle are installed, and the autonomous type that travels while finding the route that the guided vehicle should take is Applicants have applied for various patents and have been announced at various academic societies.

Further, in the field of unmanned dump trucks traveling on uneven terrain outdoors, an invention has been already known in which the traveling road surface is assumed to be a two-dimensional plane and a planned route is taught to allow the dump truck to travel autonomously. .

[0004]

However, the activity of monitoring debris flows caused by volcanic eruptions and pyroclastic flows is subject to unpredictable natural violence, and in semi-autonomous tracked vehicles as described above, the There is a risk that the members who are involved in the disaster may be caught in a disaster.

Therefore, there is a demand for the development of a safe, autonomous vehicle in which the personnel need only command and supervise at a rear point without attending.

In addition, in the automatic guided vehicle equipped with the sensor as described above, a new guidance method is proposed every year at the academic conference every time a new sensor is devised. Development speed is remarkable. Therefore, by the time the whole system was completed a few years after the development of the guidance system was started, the old sensor, which was the premise of the development, would become obsolete, and the whole system developed with much care would also become obsolete. Become. In this case, in order to avoid obsolescence, even if a newly developed high-performance sensor is to be incorporated into a system completed by combining various sensors, it is assumed that the sensor originally designed is incorporated into the entire system. Since it has been built, the entire design will have to be redone, resulting in a loss of development cost and development time.

Therefore, even if a sensor is incorporated into a vehicle for disaster relief, it is desired to develop a device which can easily incorporate a newly developed sensor in the middle of development in accordance with future development trends. There is.

[0008] In addition, the site where the autonomous disaster relief vehicle actually travels is not a flat road surface that is merely rugged like a dump truck travels, but cliffs, rivers, slopes, mountains, valleys, etc. It is a natural disaster area that has three-dimensional undulations. Moreover, since it is a natural disaster area, the terrain constantly changes, and new obstacles that were initially determined not to exist may occur. In order to guide the vehicle traveling on the road surface that changes fluidly in this way, it is not possible to directly adopt the above-described guidance method in which the traveling road surface is assumed to be a two-dimensional plane and a planned route is taught.

Therefore, there is a demand for the development of a device capable of reliably traveling in a fluidly changing natural disaster area.

The present invention has been made in view of the above circumstances, and is a running control device for an unmanned vehicle capable of safely and securely unmanned in a fluidly changing site and immediately responding to future sensor development trends. Its purpose is to provide.

[0011]

Therefore, in the main invention of the present invention, in a travel control device for an unmanned vehicle that autonomously travels along the predetermined route, data indicating the position of the predetermined route on a map is provided. Input means for inputting, and a detection means provided in the unmanned moving body for detecting an obstacle existing around the unmanned moving body with a predetermined detection range, and a position of the predetermined route inputted to the input means. First inference means for inferring an action to be taken by the unmanned vehicle based on data and the current position of the unmanned vehicle;
Second inference means for inferring an action to be taken by the unmanned moving body based on the detection result of the detection means; and the first inference means.
And a judgment means for finally judging the action based on each inference result of the second inference means, and a travel control means for controlling the travel of the unmanned vehicle based on the judgment result of the judgment means. .

[0012]

In other words, according to this structure, the data indicating the position on the map of the predetermined route is input, and the unmanned vehicle should take the data based on the position data of the predetermined route and the current position of the unmanned vehicle. Behavior is inferred. On the other hand, an obstacle existing around the unmanned moving body is detected by the detecting means within a predetermined detection range, and the above action is inferred based on the detection result. Then, the above-mentioned action is finally determined based on these inference results, and the traveling of the unmanned vehicle is controlled based on the determination results.

As described above, since the command such as the path of the moving body is comprehensively judged on the basis of the two kinds of different information such as the input data and the detection result of the sensor, the judgment function is enhanced,
Even at a fluid site, safe and reliable autonomous traveling can be performed without the need for an operator such as a member. Furthermore, since inference based on the above input data and sensor detection results is made individually, even if a newly developed sensor is incorporated during development, only the inference contents need to be changed individually, and the whole new There is no need to recreate it. Moreover, even if another sensor is added, it is only necessary to individually design and incorporate the inference content corresponding thereto, and it is possible to easily cope with future sensor development trends.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of a traveling control device for an unmanned vehicle according to the present invention will be described below with reference to the drawings. In addition, in the embodiment, it is assumed that the present invention is applied to a disaster-relief tracked vehicle that travels in a natural disaster area, but any mobile body that travels on a road surface can be used, and the application and the type of travel mechanism (wheel type , Crawler type, etc.) is optional.

FIG. 1A schematically shows the external appearance of the embodiment apparatus, and FIG. 1B conceptually shows the configuration of the embodiment apparatus in a block diagram.

A disaster relief tracked vehicle 1 shown in FIG.
(Hereinafter, simply referred to as “vehicle 1”)
The same so-called 4 as disclosed in No. 121296.
It is a tracked vehicle with a wheel truck frame structure, and the front side track 2
Also, the rear crawler belt 3 is freely rotatable, which allows the vehicle to travel while effectively gripping an uneven road surface in a natural disaster area. Then, the antenna 4 provided on the vehicle 1 and the building 1 serving as a rescue headquarters located at a location away from the natural disaster site
Transmission and reception as will be described later is performed wirelessly via the antenna 14 provided at 0.

The building 10 has a remote control device 11 which will be described later.
An input unit 13 including a monitor 12, a keyboard, and the like is provided, and an operator of the building 10 sends a command to the vehicle 1 via the monitor 12 and the input unit 13 by the wireless communication. Receive the information that is sent to you. Although wireless communication by electromagnetic waves is adopted as the communication means, any communication method such as optical communication using a protectively coated optical fiber cable can be adopted.

Now, in more detail with reference to the block diagram shown in FIG. 1B, the vehicle 1 detects the surroundings of the vehicle 1 in a predetermined detection range, that is, local detection within a radius of several meters to several hundred meters from the vehicle 1. Range (hereinafter referred to as "local vision"
That is, the sensor 5 that detects an obstacle around the vehicle 1 and a detection range different from the sensor 5, that is, the vehicle 1
To a detection range closer to the point where the radius is several tens of meters (hereinafter referred to as "near field of view").
That is, a sensor 6 for detecting an obstacle around the vehicle 1 is mounted. A CCD camera or the like for picking up the local visual field can be used as the sensor 5, and a tactile sensor, an ultrasonic sensor, a CCD camera or the like for detecting or picking up the near visual field can be used as the sensor 6. . The detection signal L of the sensor 5 and the detection signal N of the sensor 6 are input to the on-board station 8 which is a communication device.

The moving mechanism section 7 is a mechanism section which comprises a traveling motor and the like for moving the vehicle 1, and is drive-controlled based on a traveling control command signal output from the on-board station 8 which will be described later. The moving mechanism section 7 is provided with a load torque sensor, an inclinometer, and other sensors required for traveling control, which are not shown, and the detection values of these sensors are fed back to control traveling. These detected values are added to the on-board station 8 as feedback signals. On the other hand, the vehicle 1 is provided with a gyro (not shown) for detecting the current position of the vehicle body, and a detection signal indicating the current position of the vehicle is added to the on-board station 8 together with the feedback signal. The method for detecting the current vehicle position is arbitrary and may be detected by GPS, for example.

The sensor signal L input to the on-board station 8,
N and the feedback signal are subjected to predetermined signal processing and transmitted via the antenna 4 and the antenna 14 to the ground station 21 which is a transmitting device in the remote control device 11.

At the ground station 21, the transmission signal is received and subjected to predetermined signal processing, and the sensor signal L is locally inferred by the local inference unit 16
Then, the sensor signal N is output to the proximity inference unit 17, respectively. On the other hand, the feedback signal is similarly output to the integrated database 18 which is a storage device via the mechanism control unit 20 and stored and stored as data.

The input unit 13 is a map data (see FIG. 4 (a)) which will be described later and includes a route command, map information, and an operation command.
Is input through a mouse, a joystick or the like with reference to the map displayed on the monitor 12. Here, the route instruction means the vehicle 1 on the map as shown in FIG.
Of the target route L of the point sequence P0 (moving start point), P1, P2 ... Pn
When given as ..., it is a list in which registration numbers indicating these target points are arranged in order. Further, the map information is data relating to detailed information of the target points, such as the registration number of each target point Pn (n = 0, 1, 2 ...), X.
(East), Y (north) coordinate position, radius rn indicating allowable range An of passage error, height from sea level for each target point Pn, inclination angle and its direction, degree of obstacle at passage, obstacle if If there is, there is an azimuth which is considered to be avoided and attribute information including a pointer for future item addition.

The above-mentioned operation command is to instruct which operation is to be performed at which place. The contents of the operation are from the beginning to the end of the registration number of the target point for which the operation should be taken (if it is a sequential number, only the first and the last). It consists of a willingness and a pointer for expansion in preparation for adding future items. Here, the strategy will be a numerical code that indicates whether if there is an obstacle, you should forcefully follow it, if you should investigate the possibility of bypassing it, or if you decide that it is dangerous, you should give up the operation. It is expressed and the degree is expressed by the certainty factor.

The map data is input to the global inference unit 15, and based on the map data, predetermined processing is performed as described later. Further, the global inference unit 15 outputs information such as the current vehicle position on the map of the vehicle 1 and the remaining fuel amount of the vehicle 1, and the information is displayed on the monitor 12 as a computer graphics image or the like.

The global inference unit 15 advances the vehicle 1 based on the input map data, particularly the coordinate positions X and Y on the target route L and the current position detected by the gyro and stored in the integrated database 18. This is a global reasoning to be described later about the traveling command including the direction.

On the other hand, the local inference unit 16 makes a local inference for the traveling command, which will be described later, based on the input sensor signal L. On the other hand, the proximity inference unit 17 makes a proximity inference, which will be described later, on the traveling command based on the input sensor signal N.

The inference results of these three inference units are added to the integrated database 18, sequentially accumulated and recorded as the history of inference, and also output to the action judging unit 19.

In the action judging section 19, the inference sections 15 and 16 are provided.
Based on the inference results of 17 and 17, comprehensive inference to be described later is performed, and the travel command is finally determined based on the inference result. The content determined by the action determination unit 19 is transferred to the mechanism control unit 20 via the integrated database 18.
Is output as a running command to the.

The mechanism control unit 20 is mainly composed of the action determination unit 2
A travel control command signal for forward, backward, turning, posture change, etc. is generated based on the travel command determined at 0 and the feedback signal provided via the ground station 21, and this is output to the moving mechanism unit 7. It is a thing. The travel control command signal is transmitted to the on-board station 8 of the vehicle 1 via the antenna 14 of the ground station 21 and is applied to the moving mechanism unit 7 after being subjected to predetermined signal processing, whereby the travel control command signal is obtained. According to the control, the vehicle 1 is autonomously driven.

Now, the contents stored in the integrated database 18 are not limited to the map data (FIG. 4A) as shown in FIG.
Inference result of the action determination unit 19 (FIG. 7B) and the vehicle 1
There is vehicle information about the vehicle ((c) in the figure). Here, the vehicle information (FIG. 4 (c)) is roughly classified into items such as performance specifications including vehicle body size and the like, and current conditions such as vehicle current position and the like.

Although the route command, the map information, and the operation command of the map data have already been outlined, these data are basically automatically provided when the operator inputs the target route L. It is generated. However, in the attribute information, the inclination angle, the direction thereof, the degree of obstacles when passing, and the like may differ from the initial map data as the vehicle 1 advances. The local inference unit 16 and the proximity inference unit 17 determine the difference from the initial map data as described above based on the sensor signals L and N, respectively, and determine the corresponding portion of the integrated database 18 according to the determination result. Perform the process to update the data. The global inference unit 15 can again infer the course or the like on which the vehicle 1 should travel based on the updated data.

The list of target points given by the route command in the map data can be changed by the subsequent route change command. That is, after adding a new target point to be passed to the map information in the map data, insert the registration number at an appropriate position in the order of passage in the list of registration numbers for each target point in the route command. Good. For a point that is not passed due to a route change, the map information regarding the details of the point may be deleted from the integrated database, or the registration of the route command list may be canceled.

The inference results stored in the integrated database 18 include the global inference that is the inference result in the global inference unit 15, the local inference that is the inference result in the local inference unit 16, and the inference in the proximity inference unit 17. There are four types of results, that is, a near inference and an inference result obtained by the action judging section 19, and all have common data items and data formats.

Here, the common data items are forward / backward judgment, traveling speed judgment, turning radius judgment, judgment number of a target point to be passed next, judgment of posture to be taken, and caution should be taken. This is a determination of the direction in which there appears to be an obstacle, and in addition to this, a pointer for adding items in preparation for future function expansion. The data format is a format including certainty factor, which is often used as a representation method of inference results of expert systems.

The content of the actual inference result is a data format of a combination of numerical values and certainty factors as shown in FIG. 4 (b). For example, 80% to affirmative decision to turn left
In order to express the inference result that there is "confidence of", the data format is "judgment of forward / backward (+1 [code indicating left turn], +80 [confidence]") and "judgment of progress speed (+
2.0 [hour speed], +80 [confidence factor]) "and" turn radius determination (+5.6 [turn radius], +80 [confidence factor]) "
It is expressed as follows.

Below, the global inference unit 15 and the local inference unit 1
6 and each inference performed by the proximity inference unit 17 will be described more specifically with reference to FIGS. 2 and 5.

As shown in FIG. 2, the global inference unit 15 can grasp the target route L of the vehicle 1 from a global perspective, so to speak, based on the map data. According to the global view, it is necessary to calculate from the current position of the vehicle on the map and the target point on the target route L, and the course to be taken next needs to move slightly forward and then turn left to cross the bridge. If you make a right turn, it will be judged from the map that you will climb the mountain and proceed in a different direction. In this way, global reasoning is made. Therefore, in this case, the global inference result is
Affirmatively (polarity +) the judgment "turn left (the turning radius in the left turn direction is also instructed quantitatively at the same time)" and the confidence level is 80%
To output. At the same time, the determination of "go straight (instructing forward speed at the same time quantitatively)" is output affirmatively (polarity +) with a certainty factor of 40%. Further, the judgment "turn right (the turning radius in the right turn direction is also quantitatively instructed at the same time)" is output negatively (polarity-) with a certainty factor of 60%.

The local inference unit 16 can capture the surroundings of the vehicle 1 with a local field of view viewed from the vehicle 1 unlike the above global field of view. That is, assuming that the sensor 5 is a CCD camera, by processing the captured image,
You can see that there is something that seems to be a foreign body (actually a shrub) in front of it, and that there is something black shadow (actually a river) on the left side. You can also see that "something that seems to be a foreign substance cannot be found on the right side." In this way, local inference is made. Therefore, in this case, as a result of the local inference, the determination of “turn left (the turning radius in the left turn direction is also quantitatively instructed at the same time)” is negative (polarity −), and the certainty factor is 30.
Output in%. At the same time, the determination of “go straight (instructing forward speed at the same time quantitatively)” is output negatively (polarity −) with a certainty factor of 30%. Further, the determination "turn right (the turning radius in the right turn direction is also instructed quantitatively at the same time)" is output affirmatively (polarity +) with a certainty factor of 60%.

The near visual field portion 17 captures the surroundings of the vehicle 1 with a closer visual field closer than the local visual field. That is, if the sensor 6 is a tactile sensor, it can be understood from the detection result of the sensor 6 that "the foreign matter in the front is actually a shrub and does not hinder the forward movement". You can also see that you cannot find anything that seems to be a foreign substance on either the left side or the right side. In this way, proximity inference is performed. Therefore, in this case, as the proximity inference result, the determination of "turn left (instructing the turning radius in the left turn direction at the same time quantitatively)" is output affirmatively (polarity +) with a certainty factor of 60%. At the same time, affirmative judgment that "Go straight (indicate forward speed at the same time quantitatively)" (Polarity +)
In addition, it outputs with a certainty factor of 60%. Further, the determination "turn right (the turning radius in the right turn direction is also instructed quantitatively at the same time)" is output affirmatively (polarity +) with a certainty factor of 60%.

The inference results of each of the inference units 15, 16 and 17 are added to the action determination unit 19 and the inference as shown in FIG. 5 is made.

Looking at the inference results of the inference units 15, 16 and 17 shown in FIG. 2, when making a decision "whether or not to turn left", the general rule is "turn left largely on the map. 8 is a positive opinion.
It was judged with a confidence of 0%, and locally it was judged with a 30% confidence that it is better not to turn to the left because black shadows are visible in the left turn direction. The positive opinion that "there is no obstacle to proceeding to the left turn in the investigation within the range of contacting the vehicle 1" is judged with a confidence of 60%, and the local inference result is different from the other inference results. It is inconsistent with the inference result.
Similarly, with regard to the determination of whether to go straight or turn right, any one of the global inference result, the local inference result, and the close inference result is inconsistent with other inference results.

Therefore, the action judging section 19 applies the "rule for fuzzy inference" and comprehensively infers each inference result to eliminate the above contradiction.

That is, to explain the phenomenon of "turn left", the inference result relating to the event of "turn left" is input from the global inference result, the local inference result and the close inference result, and the action judgment unit 19 is first input. Is inferred in the first inference of. This first inference is based on fuzzy inference using confidence, to which “rules for inference processing input (illustrating three rules 1, 2, 3 out of many rules) are applied” Mathematical logical operations are applied to derive "intermediate inference results." Various mathematical logical operations based on fuzzy theory have been announced at academic conferences and the like and are already known, and since they have no characteristics per se, the explanation of the calculation formulas will be omitted.

At the same time, the certainty factor of the judgment about the "intermediate inference result" is also given. The intermediate inference result with this confidence is "How confident should you turn left?"
Is an intermediate judgment that, for example, “you should turn to the left while paying attention to the left side, but you cannot say that it is true (confidence xx%)”. By the way, to the action determination unit 19, other inference results regarding “go straight” and “turn right” are input, and other “postures”,
Inference results for various events such as "next target point" and "warning direction" are also input. The intermediate judgments are similarly made by fuzzy reasoning, but specific details are omitted. .

Next, with respect to the above-mentioned intermediate inference result obtained for the left turn, the intermediate inference result regarding the above-mentioned various events and other information stored in the integrated database 18, for example, operation completion of the operation instruction in the map data. In the second inference, "rules of inference for seeking output (only three rules among X, Y, and Z are shown as examples)" in the second inference together with data such as will and operation allowance time.
Is applied to perform a logical operation based on the fuzzy theory using the certainty factor. As a result, a comprehensive inference result as the action determination unit 19, that is, for example, "be careful and move forward (confidence factor + 80%)" is derived. The above-mentioned operational command is, for example, a command with the content of "scout."

Similarly, the fuzzy inference similar to that of FIG. 5 is made based on the inference results regarding the "forward" and "right turn" events.
"Assault (confidence + 10%)" as shown in Figure 2,
A comprehensive inference result such as “Stop (confidence level −10%)” is output. The comprehensive inference result is stored in the integrated database 18 in the data format shown in FIG.

When comprehensive inference is performed in this way, the result of comprehensive inference with the highest certainty, that is, "be careful and move forward"
Is finally determined as a travel command, and this travel command is output to the mechanism control unit 20.

The mechanism control unit 20 outputs a traveling control command signal including a vehicle speed command, a posture command, etc. to the moving mechanism unit 7 according to the judgment result of the action judging unit 19, and the vehicle 1 travels according to the command. Will be.

In the mechanism control unit 20, the integrated database 18 as well as the determination result from the action determination unit 19 is obtained.
The mechanism control unit 20 is also referred to by using the feedback signals from the load torque sensor, the inclinometer, and the like while also referring to the inference results of the global inference unit 15, the local inference unit 16, and the proximity inference unit 17 stored in the. We make our own judgment, for example, we make our own judgment that "there is a hole in the path and we make an emergency stop to prevent the fall," and "run an emergency stop immediately (100% confidence)" Changes are made. In this case, data such as (-99 [emergency stop], certainty factor 100%) is written in the item of the current posture of the vehicle information of the integrated database 18, and the global reasoning unit 15, the local reasoning unit 16, and the proximity reasoning unit 15. It is used as a basis for future judgments in the dynamic inference unit 17 and the action judgment unit 19.

Next, the process of changing the target route L will be described.

As described above, the map information and the route command in the map data are input by the operator in advance. However, as the vehicle 1 advances, the actual topography may be inconsistent with the map data. . Especially at natural disaster sites. Therefore, the local inference unit 1
6 and the proximity inference unit 17 determine the difference from the map data based on the input sensor signals L and N, and update the contents of the map data so as to match the current situation. By displaying the updated contents on the monitor 12, the operator can appropriately change the route L.

Similarly, regarding the operation command in the map data, the action judging section 19 judges whether the operation command is appropriate or not. For example, if it is judged that the operation command is difficult to carry out, the confidence level of the operation will be lowered. A change plan such as increasing the value of the allowable operation time can be displayed on the monitor 12 to prompt the operator to make a change. Alternatively, the action determination unit 19 itself may automatically make the change.

Further, in the case of an operation command that cannot be executed or is highly inconsistent, a command for ensuring safety such as issuing an alarm and temporarily stopping can be output, and the deterioration of the situation can be avoided. .

As described above, according to this embodiment, the vehicle 1
The travel command of is judged comprehensively based on a three-dimensional structure inference such as global inference based on the input map data and local and proximity inference based on the detection results of the sensors 5 and 6, so the determination function Therefore, even in a fluid site such as a natural disaster area, autonomous and safe driving can be performed without the need for an operator such as a member. Also, the input data and the sensor 5,
Since the inference based on the detection result of 6 is performed individually, even if the newly developed sensor is recombined during development, only the inference content needs to be changed individually, and there is no need to recreate the whole. .

In the embodiment, the case where the number of sensors is two has been described, but one of them may be omitted in some cases, and local or proximity inference may be performed based on one type of sensor. . On the contrary, three or more kinds of sensors having different detection ranges may be provided, and the inference may be individually performed based on the detection result of each sensor as in the embodiment, so that the determination function may be further enhanced. In this case, it is possible to sequentially add another newly developed sensor to the two types of sensors in accordance with future development trends, and even in that case, it is sufficient to design and incorporate the inference contents individually. Since it is not necessary to redesign the system, it is possible to easily respond to future sensor development trends, and the development cost and development time are dramatically reduced.

[0056]

As described above, according to the present invention, unmanned, safe and reliable traveling can be performed on a fluidly changing site, and it is possible to immediately respond to future sensor development trends. Development costs and development time are dramatically reduced.

[Brief description of drawings]

FIG. 1 (a) is a diagram schematically showing an appearance of an embodiment of a traveling control device for an unmanned vehicle according to the present invention, and FIG. 1 (b) is a conceptual view of a configuration of the embodiment device. It is a block diagram shown in.

FIG. 2 is a diagram used for explaining inference performed by each inference unit shown in FIG. 1 (b).

FIG. 3 is a diagram illustrating a target route along which a vehicle according to an embodiment travels.

FIG. 4 is a diagram showing stored contents of an integrated database shown in FIG. 1 (b).

FIG. 5 is a diagram used to explain the inference performed by the action determination unit shown in FIG. 1 (b).

[Explanation of symbols]

 1 tracked vehicle 5 sensor 6 sensor 7 moving mechanism section 15 global reasoning section 16 local reasoning section 17 proximity reasoning section 19 behavior determination section 20 mechanism control section

[Procedure amendment]

[Submission date] September 8, 1994

[Procedure Amendment 1]

[Document name to be amended] Statement

[Correction target item name] 0042

[Correction method] Change

[Correction content]

Therefore, the action judging section 19 applies the "rule for fuzzy inference" and comprehensively infers each inference result to eliminate the above contradiction. Regarding this kind of inference, Masaharu Mizumoto: “Various fuzzy inference methods-if ... t
In the case of hen ... ”or Mitsuru Ishizuka:“ Dempster & Sha
fer's Probability Theory ”, IEICE, 9, pp900-903 (1
The technology described in the literature such as 983) can be applied as appropriate.

[Procedure Amendment 2]

[Document name to be amended] Statement

[Correction target item name] 0043

[Correction method] Change

[Correction content]

That is, to explain the phenomenon of "turn left", the inference result relating to the event of "turn left" is input from the global inference result, the local inference result and the close inference result, and the action judgment unit 19 is first input. Is inferred by the first inference of. This first inference is applied with "rules for inference that processes input (exemplifies three rules 1, 2, and 3 out of many rules)", for example, the probability of the above-mentioned document "Dempster &Shafer". Mathematical logical operations based on fuzzy inference using certainty factor by theory are applied to derive "intermediate inference result". Here, the content of the "first inference" will be described with reference to FIG. In each inference result input, "a big turn should be made to the extreme left", "a left turn should be made as close as possible",
The item "There is no obstacle for turning left in close proximity",
It can be expressed as follows. First of all,
The respective inference units of the respective proximities infer the certainty regarding the items of "a left turn should be made", "a left turn should not be made", and "a left turn should not be made" for a left turn. For example, the degree of certainty of “should turn left” of global reasoning is 80
%, The degree of certainty that the global reasoning should not turn left is 0
%, When the certainty factor of "I don't know whether to turn left" in the global reasoning is 20%, the result of the global reasoning is 80% and the polarity of "turn left" is polar as shown in FIG. It can be easily expressed as +, and its meaning can be interpreted as "should turn extremely left". The degree of certainty of "should turn left" of local inference is 0%, the degree of certainty of "should not turn left" of local inference is 30%, and the degree of certainty of "do not know whether to turn left" of local inference When the ratio is 70%, the result of the local inference can be easily expressed as shown in Fig. 2 with the certainty factor of "turn left" is 30% and the polarity is "-". Not
Can be interpreted as Proximity reasoning for "must turn left" is 60%, proximity reasoning for "should not turn left" is 0%, and proximity reasoning for "don't know whether to turn left" confidence When it is 40%, the result of the near-field inference can be easily expressed as shown in FIG. 2 when the certainty factor of “turn left” is 60% and the polarity is +. Can be interpreted as "no." The general, local, and proximity inference units similarly infer the certainty factors of three items of “should”, “should not”, and “don't know” for straight ahead and right turn, respectively. Now,
It is assumed that the inference result as described above for the left turn is output from each of the global, local, and proximity inference units. First, in rule 1, the degree of certainty that "a global turn should be made, a local turn should be made, and a close turn should be made" is calculated, and "a large turn is made." Should be "
You can use the certainty factor. In the above example, the global reasoning “certain should turn left” is 80%, the local reasoning “should turn left” is 0%, and the proximate reasoning “should turn left”. The certainty factor of "Yes" was 60%. Here the combination that global reasoning insists "should turn left", local reasoning insist "should turn left", and proximity reasoning insist "should turn left" The certainty factor is 0.8 × 0.0 × 0.6 = 0.0. That is, the certainty factor of "should turn to the left" should be 0.0. Since the amount of information is too small, the certainty factor "I don't know whether to turn left" of the global reasoning 20
% Certainty of "I don't know if I should turn left" of local reasoning 70
% Confidence of “I don't know if I should turn left” 40 of proximity reasoning
Taking into account the percentage, at least one of the global, local, and proximity claims that they should "turn left", and none of them claim that they should "turn left", the combination is as follows. Can be calculated. 0.8 x 0.7 x 0.6 + 0.8 x 0.7 x 0.4 +
0.2 × 0.7 × 0.6 = 0.644 That is, the certainty factor of the combination that is passive but supports “must make a left turn” is 0.644. Further, the certainty factor of the combination in which the global, local, and proximity inference units all say "I do not know whether to turn left" is 0.2 x 0.7 x 0.4 = 0.056. This is a situation in which neither one agrees nor disagrees that "a left turn should be made". On the other hand, what opposes "must turn left" is a combination that claims that "no turn should be left" at any of the global, local, and proximity, and can be calculated as follows. 0.8 x 0.3 x 0.6 + 0.8 x 0.3 x 0.4 +
0.2 × 0.3 × 0.6 + 0.2 × 0.3 × 0.4 =
Summarizing 0.3 and above, we calculated the certainty factor that "a left turn should be made globally, a left turn should be made locally, and a close turn should be made", If the certainty factor of "should be", the certainty factor of "should turn to the left" is 1. Positively agree 0.0 2. If anything, support 0.644 3. I don't know which one 0.056 4. The opposite is 0.3. In this example, "must turn left" is not very strongly supported. Similarly, applying the above method to rule 2 of FIG.
Confidence that "you should turn left while paying attention to the left side" is that the global reasoning insists that "you should turn left", and the local reasoning insists that you should not turn left. You can focus on the combinations that claim to "turn left". The certainty factor of "actively agreeing" to "you should turn left while paying attention to the left side" is ◎ 0.8 x 0.3 x 0.6 = 0.144. The “opposite” confidence is 0.0. The certainty factor “I don't know which” is 0.2 × 0.7 × 0.4 = 0.056. The certainty of "somewhat supportive" is 1.0-0.144-0.056 = 0.8, and overall, if anything, "you should turn left while paying attention to your left side. "Yes" is supported. Similarly, if the above method is applied to rule 3 in FIG. 5, the certainty factor "I should not turn left for now" asserts that the global inference is "you should turn left", and the local inference is You can focus on the combinations that claim that you should turn left and that the proximate reasoning claims that you shouldn't turn left. The certainty factor of "proactively agreeing" to "don't turn left for now" is ◎ 0.8 x 0.0 x 0.0 = 0.0. The certainty of “somewhat supportive” is 0.8 × 0.7 × 0.4 = 0.224. The certainty factor “I don't know which” is 0.2 × 0.7 × 0.4 = 0.056. The certainty of "opposite" is 1.0-0.056-0.224 = 0.72, and as a whole, "I shouldn't turn left for now" is quite opposed. . Thus, the first inference is performed by applying mathematical logic operations based on fuzzy inference using confidence to the illustrated rules (rules 1, 2, and 3), for example, the following "intermediate inference". "Result" can be obtained.・ As a result of rule 1 inference, "You should turn to the left"
About 1. Positively agree 0.0 2. If anything, support 0.644 3. I don't know which one 0.056 4. Opposite 0.3 ・ As a result of rule 2 inference, "you should turn left while paying attention to the left side" 1. Positively agree 0.144 2. If anything, support 0.8 3. I don't know which one 0.056 4. Opposite 0.0 ・ As a result of rule 3 inference, "You should not turn left for now" 1. Positively agree 0.0 2. If anything, support 0.224 3. I don't know which one 0.056 4. Opposition 0.72

[Procedure 3]

[Document name to be amended] Statement

[Correction target item name] 0044

[Correction method] Change

[Correction content]

At the same time, the certainty factor of the judgment about the "intermediate inference result" is also given. The intermediate inference result with this confidence is "How confident should you turn left?"
In the above-mentioned example, the certainty factor for "I should turn left while paying attention to the left side" is 0.14.
4 (14.4%), and the certainty factor against is 0.0 (0%). By the way, the action judging section 19 is provided with another "go straight ahead",
The inference result for "turn right" is input, and the inference results for various other events such as "posture", "next target point", and "warning direction" are also input. An intermediate judgment regarding such a "right turn" and the like is made by fuzzy inference as in the case of the "left turn" described above.

[Procedure amendment 4]

[Document name to be amended] Statement

[Name of item to be corrected] 0045

[Correction method] Change

[Correction content]

Next, with respect to the above-mentioned intermediate inference result obtained for the left turn, the intermediate inference result regarding the above-mentioned various events and other information stored in the integrated database 18, for example, operation completion of the operation instruction in the map data. Along with data such as will and allowable operation time, the "inference rule for seeking output (only three rules X, Y, Z out of many rules are illustrated)" is applied in the second inference A logical operation based on the fuzzy theory using is performed. This logical operation is the first operation described above.
It can be performed in the same manner as the specific example of the "left turn" of the inference. As a result, the comprehensive inference result as the action determination unit 19,
That is, for example, "Be careful and move forward (confidence +80
%) ”Is derived. The above-mentioned operation command is, for example, a command having the content of "scout."

[Procedure Amendment 5]

[Document name to be corrected] Drawing

[Name of item to be corrected] Fig. 4

[Correction method] Change

[Correction content]

[Figure 4]

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Mitsuo Hosoi 2597, Shinomiya, Hiratsuka, Kanagawa Prefecture Komatsu Ltd. Research Department (72) Inventor Takuya Sakamoto 2597, Shinomiya, Hiratsuka City, Kanagawa Prefecture (72) Inventor Yoshihiro Naito 436-1-213 Ueki, Kamakura City, Kanagawa Prefecture (72) Inventor Takao Okui 4-3-18 Fuchinobe, Sagamihara City, Kanagawa Prefecture Nobue Corp. Room 2

Claims (2)

[Claims] 1. A travel control device for an unmanned vehicle that autonomously runs an unmanned vehicle along a predetermined route, comprising: input means for inputting data indicating a position of the predetermined route on a map; and the unmanned vehicle. A detection means for detecting an obstacle existing around the unmanned vehicle with a predetermined detection range, based on the position data of the predetermined route input to the input means and the current position of the unmanned vehicle. First inference means for inferring an action to be taken by the unmanned mobile body; second inference means for inferring an action to be taken by the unmanned mobile body based on a detection result of the detection means; And a judgment means for finally judging the action to be taken by the unmanned vehicle based on the respective inference results of the second inference means, and the traveling control of the unmanned vehicle based on the judgment result of the judging means. A travel control device for an unmanned vehicle, which comprises travel control means. 2. The detection means is at least two detection means which are provided in the unmanned moving body and which detect obstacles existing around the unmanned moving body with different detection ranges, respectively. The traveling control of the unmanned vehicle according to claim 1, wherein the inference means infers an action to be taken by the unmanned vehicle for each detection result of the detection means based on the detection results of the at least two detection means. apparatus.