CN103631261B - Information processing method and device - Google Patents
Information processing method and device Download PDFInfo
- Publication number
- CN103631261B CN103631261B CN201210309510.8A CN201210309510A CN103631261B CN 103631261 B CN103631261 B CN 103631261B CN 201210309510 A CN201210309510 A CN 201210309510A CN 103631261 B CN103631261 B CN 103631261B
- Authority
- CN
- China
- Prior art keywords
- node
- scene
- sub
- history
- present
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Abstract
The invention discloses a kind of information processing method and device.Described information processing method is applied to one or more mobile electronic equipment, and the mobile electronic equipment is used for instant positioning and the map structuring of circumstances not known.Methods described includes:N rank, wherein n >=2 are created when the instant positioning of the circumstances not known is with map structuring;In the characteristic information of the i-stage not repetition that the multiple nodes of middle filtering are included, wherein 1≤i≤n;It is single node by the multiple node stipulations.Therefore, in the present invention, this multi-level SLAM algorithm structure allows SLAM flow to be handled in a distributed manner between multiple ranks, has greatly accelerated the fusion speed of global map.
Description
Technical field
The present invention relates to field of computer technology, more particularly it relates to a kind of information processing method and device.
Background technology
Immediately positioning and map structuring(Simultaneous Localization and Mapping, SLAM)It is current
Hot research topic in terms of robot localization.So-called SLAM is exactly by mobile electronic equipment(For example, mobile robot)
Positioning with environmental map create combine together, i.e., robot in motion process according to itself pose estimate and sensors towards ambient
Perception structure increment type environmental map, while the positioning of itself is realized using the map.
When robot long-play is to carry out SLAM, the scene scale undergone constantly expands, and SLAM execution is opened
Pin will become unacceptable, can not finally meet real-time SLAM requirement, its key problem be in large scale scene how
The problem of realizing real-time SLAM.
Large scale scene is divided into multiple small-scale sub-scenes by existing first solution according to geographical space, to every
Individual sub-scene carries out real-time SLAM, and multiple sub-scenes of most at last SLAM generations merge into global large scale scene.
However, in this fashion, with the switching of sub-scene, SLAM will restart completely, two SLAM sequentially
It is constructed go out sub-scene map can produce larger deviation, have a strong impact on the degree of accuracy of final global map.
Thus, people are further contemplated that can be phase by above-mentioned first solution and by way of carrying out more scene switchings
With reference to the second solution realize real-time SLAM.
However, due to bulk redundancy(overlapping)Characteristic information and cartographic information, result in matching largely include
The small-scale scene of duplicate message also becomes extremely difficult and slow.Moreover, when occurring loop in environmental map, need
Quickly to identify and registering with scene progress before, so as to obtain consistent global map.
For problem above, in the prior art, further 3rd solution is increased again in existing SLAM flows
Add a rank(level)SLAM flows, i.e., the SLAM flows of two ranks, for realize the loop detection of global map and
Map merges.
But with high-level(high level)Nodal information is continuously increased during SLAM, necessarily also results in it
The characteristic information of redundancy increases with cartographic information, so that high level fusion process matching error increases, speed is slow, and then
Have a strong impact on the degree of accuracy of final global map.
The content of the invention
In order to solve the above-mentioned technical problem, according to an aspect of the invention, there is provided a kind of information processing method, application
In one or more mobile electronic equipment, the mobile electronic equipment is used for instant positioning and the map of circumstances not known
Structure, it is characterised in that methods described includes:N rank is created when the instant positioning of the circumstances not known is with map structuring,
Wherein n >=2;In the characteristic information of the i-stage not repetition that the multiple nodes of middle filtering are included, wherein 1≤i≤n;Will be the multiple
Node stipulations are single node.
In addition, according to another aspect of the present invention, there is provided a kind of information processor, can applied to one or more
Mobile electronic device, the mobile electronic equipment are used for instant positioning and the map structuring of circumstances not known, it is characterised in that
Described device includes:Level creation unit, for creating n level when the instant positioning of the circumstances not known is with map structuring
Not, wherein n >=2;Characteristic filter unit, for the characteristic information in the i-stage not repetition that the multiple nodes of middle filtering are included, its
In 1≤i≤n;Node stipulations unit, for being single node by the multiple node stipulations.
Compared with prior art, using the information processing method and device according to the present invention, in robot to circumstances not known
When performing positioning immediately and map structuring, multiple ranks are created(level)SLAM algorithm structures, to it in each rank
In the characteristic point information of redundancy that is included of each node and cartographic information filtered, and by multiple node stipulations after filtering
For single node.Therefore, in the present invention, this multi-level SLAM algorithm structure allows SLAM flow in multiple levels
Handled in a distributed manner between not, greatly accelerated the fusion speed of global map.
Other features and advantages of the present invention will be illustrated in the following description, also, partly becomes from specification
Obtain it is clear that or being understood by implementing the present invention.The purpose of the present invention and other advantages can be by specification, rights
Specifically noted structure is realized and obtained in claim and accompanying drawing.
Brief description of the drawings
Accompanying drawing is used for providing a further understanding of the present invention, and a part for constitution instruction, the reality with the present invention
Apply example to be used to explain the present invention together, be not construed as limiting the invention.In the accompanying drawings:
Fig. 1 a and 1b illustrate the principle of the solution according to prior art.
Fig. 2 illustrates the information processing method according to the present invention.
Fig. 3 illustrates the information processor according to the present invention.
Fig. 4 illustrates the principle of information processing method according to embodiments of the present invention.
Fig. 5 illustrates information processor according to embodiments of the present invention.
Fig. 6 illustrates the information processing method in the 1st rank according to embodiments of the present invention.
Fig. 7 illustrates the information processing method in intermediate level according to embodiments of the present invention.
Fig. 8 illustrates the information processing method in the n-th rank according to embodiments of the present invention.
Fig. 9 a illustrate the cartographic model of the circumstances not known for experimental demonstration.
Fig. 9 b to Fig. 9 l illustrate the view data that robot is shot in circumstances not known.
Figure 10 a to Figure 10 b respectively illustrate the effect mould of the map generated according to the 3rd solution of prior art
The top view and side view of type.
Figure 11 a to Figure 11 b illustrate the effect mould for the map that information processing method according to embodiments of the present invention is generated
The top view and side view of type.
Embodiment
It will be described in detail with reference to the accompanying drawings each embodiment according to the present invention.Here it is to be noted that it in the accompanying drawings,
Identical reference, which is assigned, substantially has the part of same or like 26S Proteasome Structure and Function, and will omit on it
Repeated description.
First, the principle of the solution according to prior art will be described with reference to figure 1a and 1b.
Fig. 1 a and 1b illustrate the principle of the solution according to prior art.
In the prior art, in order to build the environmental map of circumstances not known, mobile electronic equipment(For example, mobile machine
People)Constantly moved in circumstances not known, obtain each location point #1, #2, #2 ', #3, #3 ', #4 and #4 ' etc. image are seen
Survey data, positional information and attitude information, judge these data messages whether for describe the circumstances not known whether be it is effective,
And utilize effective data message(For example, #1, #2, #3 and #4 etc.)To carry out SLAM, passed through so as to obtaining robot
Track x1, x2 and x3 etc., and finally obtain the environmental map.
When circumstances not known has large scale scene, in order to reduce the executive overhead in large scale scene, first solves
Scheme proposes whole large scale scene being divided into multiple small-scale sub-scene M1, M2 and M3 etc., to every height in robot
Scene independently carries out SLAM, for example, after being finished on the SLAM operations of first sub-scene, will be on first son
On the sub- map of scene is uploaded onto the server, and related data message is deleted from robot, then the weight in robot
The new SLAM operations for starting second sub-scene, and finally on the server piece together more sub- maps on circumstances not known
Global map, as illustrated in Fig. 1 a.
For the offset issue between two continuous sub-scene maps caused by reducing the first solution, the second solution party
Case proposes that the SLAM for just starting to start second sub-scene is operated at the end of the SLAM operations not yet of first sub-scene, from
And generate and include redundant data(For example, data message between sub-scene M1 and M2 at the location point #2 of redundancy, in sub-scene
Data message between M2 and M3 at the location point #3 of redundancy etc.)Sub- map, these redundant datas can improve sub- map
Precision, map error caused by sub-scene switching is effectively reduced, while the deviation for merging global map will be reduced.
Further, in order to solve in the second solution due to largely causing matching small-scale comprising duplicate message
Scene also becomes extremely difficult problem, and for the loop occurred in environment-identification map, the 3rd solution(Also referred to as
Mix rank SLAM principles(Hybrid level SLAM Principal))It is proposed:Circumstances not known is performed in the mobile robot
SLAM when create two ranks, i.e. local level(local level)And global level(global level), in low level
(That is, local level)In realize the second solution, and high-level(That is, global level)The middle loop for realizing global map
Detection and map fusion, as illustrated in Fig. 1 a and 1b.
Although the 3rd solution solves many problems of the prior art, with constantly prolonging for SLAM processes
Continuous, the nodal information during global level SLAM can be also continuously increased, so as to necessarily cause vision guide(Visual-
Guided)SLAM will generate a large amount of identical features, and ultimately cause substantial amounts of error hiding.
On the other hand, present inventors have proposed a kind of new information processing method and device.
Hereinafter, by with reference to figure 2 and 3 come describe according to the present invention information processing method and device.
Fig. 2 illustrates the information processing method according to the present invention, and Fig. 3 illustrates the information processing apparatus according to the present invention
Put.
Information processing method illustrated in Fig. 2 is applied to one or more mobile electronic equipment, the removable electricity
Sub- equipment is used for instant positioning and the map structuring of circumstances not known.Also, described information processing method passes through illustrated in Fig. 3
Information processor is realized.Specifically, described information processing unit 100 includes:Level creation unit 110, characteristic filter list
Member 120 and node stipulations unit 130.
As illustrated in Figure 2, described information processing method includes:
In step s 110, level creation unit 110 creates n in instant positioning and the map structuring of the circumstances not known
Individual rank, wherein n >=2;
In the step s 120, the feature for the repetition that characteristic filter unit 120 is included in the not middle multiple nodes of filtering of i-stage
Information, wherein 1≤i≤n;
In step s 130, the multiple node stipulations are single node by node stipulations unit 130.
As can be seen here, using the information processing method and device according to the present invention, it is to circumstances not known execution in robot
When Shi Dingwei is with map structuring, multiple ranks are created(level)SLAM algorithm structures, to wherein each in each rank
The characteristic point information and cartographic information for the redundancy that node is included are filtered, and are single by multiple node stipulations after filtering
Node.Therefore, in the present invention, this multi-level SLAM algorithm structure allows SLAM flow between multiple ranks
Handled in a distributed manner, greatly accelerated the fusion speed of global map.
Hereinafter, information processing method and device according to embodiments of the present invention will be described with reference to figure 4 to Fig. 5.
Fig. 4 illustrates the principle of information processing method according to embodiments of the present invention, and Fig. 5 is illustrated according to of the invention real
Apply the information processor of example.
Information processing method according to embodiments of the present invention illustrated in Fig. 4 can apply at the information illustrated in Fig. 5
Manage device 100.As illustrated in fig. 5, the information processor 100 includes:Level creation unit 110, characteristic filter unit 120
With node stipulations unit 130.Moreover it is preferred that the information processor 100 also includes:Scene creation unit 140, node obtain
Take unit 150, map constructing unit 160, node adding device 170 and processing unit 180.
As illustrated in Fig. 2, Fig. 4 and Fig. 5, in information processing method according to embodiments of the present invention, when removable electricity
Sub- equipment(For example, mobile robot)When being used for the instant positioning of circumstances not known with map structuring, level creation unit 110 is first
N rank, wherein n >=2 are first created according to user's setting.
For i-stage is other, wherein 1≤i≤n, scene creation unit 140 is in i-stage not the i-th scene of middle establishment, section
Point acquiring unit 150 is from the i-th -1 rank or the camera by being equipped in robot is believed seriatim to obtain characteristic point
Breath and cartographic information, map constructing unit 160 carry out SLAM corrections according to these information, created on the scene partly
Figure.Also, when judging that present node is effective, node adding device 170 is added to effective history node.Then,
When the node obtained in the other interior joint acquiring unit 150 of i-stage increases to certain number, i.e., when i-stage not in section
When the information and excessive cartographic information of the feature that point includes, characteristic filter unit 120 is believed identical characteristic point in the node
Breath and cartographic information are filtered, and multiple node stipulations after filtering are single node by node stipulations unit 130.Most
Afterwards, the single node is sent to i+1 rank by processing unit 180, and i-stage not in filtering after the multiple section
Among point, only retain last predetermined number node, delete other nodes, to continue the other positioning immediately of i-stage
With map structuring.
Recursively, in the i+1 rank carried out with the i-th level parallelism, node acquiring unit 150 is constantly from i-stage
The not middle single node obtained after stipulations, and create the local map on i+1 rank.Similarly, when node therein is believed
Breath is when exceeding certain threshold value, by characteristic filter unit 120 and node stipulations unit 130 it filter simultaneously stipulations to i-th+2
In the nodal information of rank, so repeatedly, until the final node in the n-th rank by store without any redundant data
Whole characteristic informations and cartographic information, that is, form comprising the finish node on whole circumstances not known information.
Below, by with reference to figure 5 and Fig. 6 to Fig. 8 come the letter in describing i-stage according to embodiments of the present invention in detail not
Processing method is ceased, wherein 1≤i≤n.
First, the information processing method that will be described with reference to Figure 6 in the 1st rank according to embodiments of the present invention.
Fig. 6 illustrates the information processing method in the 1st rank according to embodiments of the present invention.
As illustrated in FIG. 6, the 1st rank according to embodiments of the present invention(That is, i=1)In information processing method using double
The SLAM methods of graphic structure realize that the principle of the SLAM methods of the dual graphic structure is:In a scene SLAM not yet
At the end of just start to start next scene SLAM, i.e. perform dual graphic SLAM(Duo-Graph SLAM).Therefore, based on
The transitional period that the SLAM of scene is switched over, it will there are two separate scenes carrying out SLAM simultaneously.
Specifically, the information processing method in the 1st rank includes:
In step s 200, this method starts to perform.
When people need to build map to a circumstances not known, mobile robot is used for the SLAM for carrying out circumstances not known.
At this moment, level creation unit 110 creates n rank according to user's setting first, and wherein n is natural number, and n >=2.So
Afterwards, the notification scenario creating unit 140 of level creation unit 110 starts to create the SLAM figures of the circumstances not known in the 1st rank
Structure, i.e. the 1st scene.
In step S205, frame observation data are obtained.
Specifically, scene creation unit 140 creates the 1st scene in the 1st rank.Due to using digraph in the 1st rank
Shape SLAM technical scheme, i.e. the 1st scene actually includes two continuous scenes, so in order to distinguish, will be in the 1st scene
Two scenes included are referred to as the first sub-scene and the second sub-scene.
So scene creation unit 140 creates the first sub-scene of the 1st scene first.Then, as mobile robot exists
Move and shoot in circumstances not known, information processor 100 obtains image observation data in each location point, robot
Own location information and attitude information.Specifically, node acquiring unit 150 is in the first sub-scene of the 1st rank created
Frame by frame obtains above-mentioned data message.For example, it obtains the first frame observation data, the first node as the first sub-scene first
N101.First frame observation data can be including image observation data of the mobile robot in first position point, robot itself
Positional information and attitude information etc..
In step S210, local map structure is carried out according to effective historical frames of present frame and the first sub-scene.
Specifically, after node acquiring unit 150 gets the first frame observation data, map constructing unit 160 performs
Operation, to carry out local map structure to the present node included in the first sub-scene and history node.Due to now existing
Only include first node N101 in first sub-scene, so map constructing unit 160 individually enters to first node N101
Row local map is built.
In step S215, judge whether to meet the first predetermined condition.
Specifically, node adding device 170 in map constructing unit 160 to the present node that includes in the first sub-scene
After carrying out local map structure with history node, whether judgement currently meets the first predetermined condition.For example, first here is pre-
Fixed condition refers to that the characteristic point of present frame is enough, and the office carried out according to effective historical frames of present frame and the first sub-scene
Portion's map structuring success.
When node adding device 170 judges that the characteristic information of the present node is enough and in the first subfield
When present node and history node the progress local map that scape includes successfully construct, this method proceeds to step S220.It is another
Aspect, if determining that present frame is unsatisfactory for the first predetermined condition in step S215, this method returns to step S205, with weight
It is new to obtain frame observation data, such as the second frame observation data.
In step S220, present frame is added as to effective historical frames of the first sub-scene.
Specifically, the first frame observation data that node adding device 170 gets node acquiring unit 150, i.e., first
Node N101, it is added to effective history node.
In step S225, judge whether to meet the second predetermined condition.
Next, scene creation unit 140 judges whether the first sub-scene meets the second predetermined condition.Here second is pre-
Fixed condition is the hardware configuration limitation due to the mobile electronic equipment, otherwise described removable except the non-clear data
Electronic equipment can not be further continued for carrying out SLAM.
Specifically, scene creation unit 140 for example may determine that the present node included in the first sub-scene and history
Whether the total number of node is more than or equal to first threshold.If meeting this Rule of judgment, this method proceeds to step S230;
Otherwise, this method returns to step S205.
Here, suppose that first threshold is 6.Because the node now included in the first sub-scene is only first node
N101, i.e. node total number are only 1, so and be unsatisfactory for the second predetermined condition, then this method returns to step S205,
To reacquire frame observation data, such as the second frame observation data.
So repeatedly, until node acquiring unit 150 gets the 6th frame observation data, the as the first sub-scene
Six node N106.At this moment, map constructing unit 160 is to the present node that includes in the first sub-scene(That is, the 6th node
N106)With history node(That is, first node N101 to the 5th node N105)Carry out local map structure.If meet that first is pre-
Fixed condition, then node adding device 170 the 6th node N106 is added to effective history node of the first sub-scene.Then,
When scene creating unit 140 judge the total number of the present node and history node included in the first sub-scene whether be more than or
During equal to first threshold, it will find that total number 6 has equalized to first threshold 6, then this method proceeds to step S230.
In step S230, the second sub-scene is judged whether.
Scene creation unit 140 determines whether currently to whether there is the second sub-scene in the 1st rank.If it is,
This method proceeds to step S235;Otherwise, this method proceeds to step S240.
In step S235, present frame is added as to effective historical frames of the second sub-scene.
Specifically, if it find that currently having created the second sub-scene, then scene creation unit 140 notifies node addition
Present node N106 is further added to the history node of the second sub-scene by unit 170.
In step S240, the second sub-scene is created, and present frame is added as to effective historical frames of the second sub-scene.
Specifically, if it find that there is currently no the second sub-scene, then scene creation unit 140 creates the of the 1st scene
Two sub-scenes, to perform dual graphic SLAM operations in the 1st rank, and notify node adding device 170 by present node
N106 is further added to effective historical frames of the second sub-scene.
In step S245, judge whether to meet the 3rd predetermined condition.
Characteristic filter unit 120 judges whether the second sub-scene meets the 3rd predetermined condition.For example, the 3rd here is predetermined
Condition is that the second sub-scene possesses sufficiently high uniformity with the first sub-scene, and the second sub-scene has been present enough
More node, so as to only carry out local map structure with the second sub-scene.
Specifically, whether the 3rd predetermined condition meets by characteristic filter unit 120 to judge in the second sub-scene
Including present node and the total number of history node whether be more than or equal to Second Threshold and realize.If meet that the 3rd is predetermined
Condition, then this method proceed to step S250;Otherwise this method returns to step S205, and number is observed to reacquire a new frame
According to.
Here, for convenience, it is assumed that Second Threshold is 2.Then, when node acquiring unit 150 gets the observation of the 7th frame
Data, during the 7th node N107 as the first sub-scene, characteristic filter unit 120 judges what is included in the second sub-scene
Present node and history node are N106 and N107, i.e., total number is 2, has equalized to Second Threshold, then this method proceeds to
Step S250.
In step s 250, duplicate message is filtered, stipulations transmission single node, the second sub-scene is switched into the first subfield
Scape, delete the second sub-scene.
When the present node and the total number of history node judging to include in the second sub-scene reach Second Threshold,
Characteristic filter unit 120 filters the present node included in the first sub-scene and history node(That is, N101 to N107)Wrapped
The characteristic information of the repetition contained.
Specifically, characteristic filter unit 120 is from first node N101 to the 7th node N107 observation data
(observation)Middle extraction characteristic information(feature)And cartographic information, judge with the presence or absence of repetition to believe between these information
Breath.If it is, characteristic filter unit 120 filters to the characteristic information and cartographic information that repeat, so as to reduce the 1st
Rank performs the expense and redundancy during SLAM.
The spy for the repetition that present node and history node are included is filtered out in the 1st rank in characteristic filter unit 120
After reference breath, multiple node stipulations after filtering are single node by node stipulations unit 130.Preferably, feature is being carried out
During filtering, node stipulations unit 130 also carries out strengthening Map recognition operation to each characteristic information, to reduce in spy
Error that may be present in reference breath, and improve the precision of Map recognition.
Specifically, the characteristic information difference for the repetition that node stipulations unit 130 is included to present node and history node
Weighted value is assigned, for example, the allocation rule of weighted value can be:If robot is nearer apart from the distance of this feature point, weigh
Weight values are bigger.Here, suppose that the disc floret that this feature point is robot to be photographed during traveling, and robot is first
Apart from 9 meters of this disc floret in node, apart from 6 meters of this disc floret in section point, apart from 3 meters of this disc floret in the 3rd node, then may be used
To assign weighted value 5 to first node, weighted value 6 is assigned to section point, weighted value 7 is assigned to the 3rd node.Then, node
Stipulations unit 130 is weighted averagely according to the weighted value come the characteristic information of the repetition included to multiple nodes, and
The single node of the characteristic information after including weighted average is generated, and finally deletes the multiple node.
Next, processing unit 180 node stipulations unit 130 by the multiple node stipulations be single node after,
It is discarded in all nodes that the first sub-scene includes(That is, first node N101 to the 7th node N107), and with the second son
Scene continues the SLAM in the 1st rank.For example, processing unit 180 is removed simultaneously using the second sub-scene as the first sub-scene
Second sub-scene is sold, to continue the instant positioning of the 1st rank and map structuring.
Finally, the processing unit 180 by the single node that node stipulations unit 130 is generated from current level(That is, the 1st
Rank)It is sent to higher level(That is, the 2nd rank), to carry out the instant positioning of the 2nd rank and map structuring.
For example, in the new round iteration of SLAM in the 1st rank, the first new son for coming from the switching of the second sub-scene
Currently the 6th node N106 and the 7th node N107 is only included in scene.Node acquiring unit 150 obtains continuation frame by frame and obtained
Observation data are taken, when node acquiring unit 150 gets ten one node of the 11st frame observation data as the first sub-scene
During N111, the total number for the node that the first sub-scene includes is equal to first threshold 6 again.Then, scene creation unit 140
The second sub-scene is created again, and effective 11st node N111 is added to effective history node of the second sub-scene.
Next, when node acquiring unit 150 gets the 12nd frame observation data, the 12nd node N112 as the first sub-scene
When, characteristic filter unit 120 judges the present node included in the second sub-scene and history node is N111 and N112, i.e.,
Total number is 2, has equalized to Second Threshold, and then this method filters the duplicate message of all nodes in the first sub-scene again,
Stipulations simultaneously transmit single node to the 2nd rank, the second sub-scene are switched into the first sub-scene, and delete the second sub-scene.
So circulate repeatedly, untill periods of robot operation stop.
It should be noted that the SLAM operations of the 1st rank can preferably be completed in robot.Furthermore, it is possible to work as
While robotic end abandons the data of the first scene, the data of discarding are stored in long-range host end(For example, server),
For subsequent operation.
As can be seen here, using the information processing method in the 1st rank according to embodiments of the present invention, i.e. Duo-Graph
SLAM working method, the sub- map comprising redundant data can be generated, these redundant datas can improve the precision of sub- map,
Map error caused by the switching of the SLAM based on sub-scene is effectively reduced, while the deviation for merging global map will be reduced.
In addition, this Duo-Graph SLAM working method can also continuously carry out SLAM, found without for good and all losing
Map.In addition, robot may be reused its own SLAM results and other robot obtained in different sessions
SLAM results.Herein on basis, establish large-scale three-dimensional(3D)Map becomes feasible.Finally, when the node in the 1st rank
When information exceedes certain threshold value, it can be filtered and stipulations are into the nodal information of the 2nd rank, so that the 2nd rank section
Put the whole characteristic informations and cartographic information of in store any redundant data without in the 1st rank.
Next, it will come with reference to figure 7 continuing on the information processing method in intermediate level according to embodiments of the present invention.
Fig. 7 illustrates the information processing method in intermediate level according to embodiments of the present invention.
As illustrated in figure 7, intermediate level according to embodiments of the present invention(That is, 1<i<n)In information processing method profit
Realized with the SLAM methods of single graphic structure, the principle of the SLAM methods of single graphic structure is:In current scene SLAM
At the end of, the last predetermined number node in the scene is retained, then start to start scene SLAM again, i.e. perform list
Figure SLAM(single-Graph SLAM).With dual graphic SLAM differently, two are not had in free hand drawing shape SLAM mutually
Independent scene is carrying out SLAM simultaneously.But, due to being remained in next scene SLAM. with dual graphic SLAM similarly
One or more nodes of a upper scene, so this free hand drawing shape SLAM modes still can be generated comprising redundant data
Sub- map, these redundant datas can improve the precision of sub- map, effectively reduce map error, while merge reducing globally
The deviation of figure.
Below, the information processing method in intermediate level according to embodiments of the present invention is explained by taking i=2 as an example.
Specifically, the information processing method in the 2nd rank includes:
In step S300, this method starts to perform.
After level creation unit 110 creates n rank according to user's setting in step s 200, level creation list
First 110 notification scenario creating units 140 start with preceding i-1 level parallelism, in the not middle establishment circumstances not known of i-stage
SLAM graphic structures, i.e. the i-th scene.In this example, it is described due to following by taking i=2 as an example, so the i-th scene is actually
For the 2nd scene.
In step S305, judge whether to get the new node of the i-th -1 rank.
Specifically, scene creation unit 140 creates the 2nd scene in the 2nd rank.Then, node acquiring unit 150 judges
Whether from the 1st rank new node data is got.If it is, node acquiring unit 150 makees the new node data
For the first node N201 of the 2nd scene.The node can include filtering the image observation number after simultaneously stipulations in the 1st rank
According to, robot own location information and attitude information etc..If not getting new node data, this method returns to
Step S305, the new node data come is caught from the 1st rank until getting.
In step S310, local map structure is carried out according to present node and effective history node of the i-th scene.
With step S210 similarly, after node acquiring unit 150 gets first node data, map constructing unit
160 perform operation, to carry out local map structure to the present node included in the 2nd scene and history node.Due to this
Only include first node N201 in the scenes of Shi 2, so map constructing unit 160 individually enters to first node N201
Row local map is built.
In step S315, judge whether to meet the first predetermined condition.
Node adding device 170 judges whether current present node meets the first predetermined condition.For example, node adding device
170 judge whether the characteristic information of the present node enough, and judge to the present node that includes in the 2nd scene and
History node carries out whether local map structure succeeds.If it is, this method proceeds to step S320;Otherwise, before this method
Step S330 is entered, is single node by the 2nd scene stipulations to filter out the duplicate message that all nodes include in the 2nd scene,
And it is other that the single node is sent into 3rd level, to carry out 3rd level other positioning and map structuring immediately.
Here, when judging that current present node is unsatisfactory for the first predetermined condition, this method is not to return to step
S305 but be the reason for directly carry out filtering stipulations:Because robot tends not to enter wrong road in the 2nd rank
Footpath, occur the situation of being held as a hostage so being likely to the robot the reason for cause present node invalid or artificially closed by operator
Break and be moved in another circumstances not known different with current circumstances not known and continue to operate, so now, generally requiring to before
The environmental map of structure is generated and shown.
It is apparent that in practice, it can also to work as according to the real needs of operator and judge currently to work as prosthomere
When point is unsatisfactory for the first predetermined condition, this method returns to step S305, and reacquires node.
In step s 320, present node is added as to effective history node of the i-th scene.
Specifically, the first node data that node adding device 170 gets node acquiring unit 150, i.e. first segment
Point N201, the effective history node being added in the 2nd scene.
In step S325, judge whether to meet the second predetermined condition.
Next, characteristic filter unit 120 judges currently whether meet the second predetermined condition.For example, characteristic filter unit
120 judge whether the total number of the present node and history node included in the 2nd scene is more than the 3rd threshold value.If meet the
Two predetermined conditions, then this method proceed to step S330;Otherwise this method returns to step S305.
Here, for convenience, it is assumed that the 3rd threshold value is 2.Then, when node acquiring unit 150 is got from the 1st rank
Second single node coming up is transmitted, during section point N202 as the 2nd scene, characteristic filter unit 120 is judged
The present node and history node that 2nd scene includes are N201 and N202, i.e., total number is 2, has equalized to Second Threshold, in
It is that this method proceeds to step S330.
In step S330, duplicate message is filtered, it be single node that stipulations, which transmit the i-th scene, reservation destined node, by the
Other nodes of i scenes empty.
With step S250 similarly, when the present node and the total number of history node judging to include in the 2nd scene
During more than three threshold values, characteristic filter unit 120 and node stipulations unit 130 are by the first node N201 in the 2nd scene
Performed with the characteristic information in section point N202 observation data and cartographic information and remove redundancy and enhancing Map recognition
Operation, to filter the repeated characteristic information that all nodes are included in the 2nd scene, and all nodes in the 2nd scene are advised
About single node.
Processing unit 180 is after the multiple node stipulations are single node by the node stipulations unit 130, the
Retain last predetermined number destination node in 2 scenes, to continue the instant positioning of the 2nd rank and map structuring, delete it
Remaining node, and it is other that the single node is sent into 3rd level, to carry out 3rd level other positioning and map structuring immediately.
Specifically, in Fig. 4 illustrated examples, the predetermined number is 1.That is, processing unit 180 is by the 2nd
First node N201 in scape is deleted, and retains section point N202, and backs within step S305, to obtain Section three
Point N203, so as to continue the instant positioning of the 2nd rank and map structuring.
It should be noted that the SLAM operations of intermediate level can preferably be completed on the server, to utilize service
The memory space of device magnanimity and the computing capability of high speed conveniently and efficiently build local map data.In addition, above-mentioned 3rd threshold
Value, which is the other SLAM figures of i-stage, carries out the upper limit thresholds of stipulations, the data included due to different ranks to SLAM next time
Amount is different, so the 3rd threshold value in each rank can also be different.
As can be seen here, using the information processing method in intermediate level according to embodiments of the present invention, i.e. free hand drawing shape SLAM
Working method, still can i-stage it is not middle generation comprising be used for improve the accuracy of map redundant data sub- map, and
Finally, when i-stage not in nodal information exceed certain threshold value when, can be filtered and stipulations to i+1 rank node
In information so that i+1 level node by it is in store without i-stage not in any redundant data whole features believe
Breath and cartographic information.
Finally, will come with reference to figure 8 continuing on the information processing method in highest level according to embodiments of the present invention.
Fig. 8 illustrates the information processing method in the n-th rank according to embodiments of the present invention.
As illustrated in Figure 8, highest level according to embodiments of the present invention(That is, i=n)In information processing method original
Reason is:Only only maintained in the n-th rank and show a node Nn01.
Specifically, the information processing method in the n-th rank includes:
In step S400, this method starts to perform.
In step S405, judge whether to get the new node of the (n-1)th rank.
In step S410, global map structure is carried out according to present node and effective history node of the n-th scene.
In step S415, judge whether to meet the first predetermined condition, if it is, this method proceeds to step S420;
Otherwise, this method, which returns, performs step S405.
In the step s 420, present node is added as to effective history node of the i-th scene.
Step S400 to the S420 is similar to step S300 to S320, in order to simple, only describes its difference:
After level creation unit 110 creates n rank according to user's setting in step s 200, level creation list
First 110 notification scenario creating units 140 start, the circumstances not known to be created in the n-th rank with preceding n-1 level parallelism
SLAM graphic structures, i.e. the n-th scene.Node acquiring unit 150 will get new node data as from the (n-1)th rank
The single node Nn01 of n scenes.Map constructing unit 160 carries out global map structure to single node Nn01 and history node
Build.When the characteristic information for judging the present node is enough, and judge the present node to including in the n-th scene
When being successfully constructed with history node progress global map, first node Nn01 is added in the n-th scene by node adding device 170
Effective history node.
In step S425, duplicate message is filtered, the scene of stipulations n-th is single node.
Characteristic filter unit 120 and node stipulations unit 130 pass through the observation number to the first node Nn01 in the n-th scene
Characteristic information and cartographic information in, which perform, removes redundancy and enhancing Map recognition operation, to filter institute in the n-th scene
There is the repeated characteristic information that node is included, and be single node by the node stipulations in the n-th scene.
In step S430, single node is shown.
Processing unit 180 shows the list after the node stipulations are single node by the node stipulations unit 130
One node, i.e., the constructed environmental map on circumstances not known is shown to user.
It should be noted that the SLAM operations of highest level can preferably be completed on the server, to utilize service
The memory space of device magnanimity and the computing capability of high speed conveniently and efficiently build global map data.
As can be seen here, can will be without using the information processing method in highest level according to embodiments of the present invention
Whole characteristic information and cartographic information stipulations of any redundant data are single node, and show the single node institute to user
Comprising global map image.
Below, by with reference to figure 9a to Figure 11 b come illustrate according to the solution of prior art with according to embodiments of the present invention
The effect of map that is generated of information processing method compare.
Fig. 9 a illustrate the cartographic model of the circumstances not known for experimental demonstration, and Fig. 9 b to Fig. 9 l illustrate robot and existed
The view data shot in circumstances not known.
As illustrated in Fig. 9 a, robot advances in the circumstances not known of square configuration, and the circumstances not known has 21 meters × 17
The length and width of rice.The robot is sequentially in time from the 1st location point(As represented by wherein including the triangle of word " 1 "
's)It is moved to the 11st location point(As represented by wherein including the triangle of word " 11 "), wherein triangular representation machine
People, and the drift angle of triangle represents the orientation pointed by robot.Robot captures surrounding environment in this 11 location points
View data, as illustrated in Fig. 9 b to Fig. 9 l.
Figure 10 a to Figure 10 b respectively illustrate the effect mould of the map generated according to the 3rd solution of prior art
The top view and side view of type, and Figure 11 a to Figure 11 b illustrate what information processing method according to embodiments of the present invention was generated
The top view and side view of the effect model of map.
As illustrated in Figure 10 a and Figure 10 b, the map that is generated according to the 3rd solution of prior art obviously with figure
The circumstances not known of square configuration illustrated in 9a exist it is dramatically different, as described above, this is due to be wrapped in the 3rd solution
Characteristic information and cartographic information containing bulk redundancy, and because a large amount of similar visual signature information cause when generating map
Substantial amounts of erroneous matching, so the map for causing to ultimately generate can not close, and produce distortion.
However, as illustrated in Figure 11 a and Figure 11 b, can when using information processing method according to embodiments of the present invention
The map ultimately generated with finding out is extremely approximate with the circumstances not known of the square configuration illustrated in Fig. 9 a, wherein detecting robot
The loop of running orbit, and form the good closure of map.
As can be seen here, using information processing method according to embodiments of the present invention(That is, multi-level SLAM algorithm structures), can
SLAM corrections are carried out with the nodal information between same rank, whenever the node of a rank increases to certain number, i.e.,
Comprising feature information and excessive cartographic information when, it is as illustrated in figure 4, identical to filter out with regard to once being filtered
Characteristic point information and cartographic information, the other multiple nodes of i-stage(In square frame shown in Fig. 4)Information is by stipulations to i+1 level
In other single node information.Equally, also can be by its stipulations to i-th when i+1 rank interior joint information exceedes certain threshold value
In+2 level node information, so circulate repeatedly.So final n-th level node by store without any redundant data
Whole characteristic informations and cartographic information.
Therefore, the algorithm of this multi-level SLAM structures not only allows SLAM flow to be handled in a distributed manner, greatly
It is big to accelerate global map fusion speed, and loop can be detected in any rank, so as to which rapid build is uniformly convergent
Global map.
It should be noted that wrapped although filtering will be used in the above example of the present invention in the other scene of i-stage
The duplicate message of all nodes included and it is illustrated as in i-stage not middle realization the step of be single node by their stipulations(Example
Such as, realized in step S250, S330 and/or S425), but the invention is not restricted to this.But can be with the contrary, will be i-th
All nodes that the scene of rank includes all are sent in i+1 rank, and in i+1 rank, first in i-stage
The duplicate message for all nodes that other scene includes is filtered, and is single node by their stipulations(For example, in step
Above-mentioned steps are realized in rapid S305 and/or S405), remaining step in i+1 rank is then performed again.
Further, it should be noted that although hereinbefore information processing method according to embodiments of the present invention is described as:
Information processing method in 1st rank is completed in robot and by middle and final information processing method on the server
Complete, but the invention is not restricted to this, the information processing method of all rank can also be all arranged in robot or service
Completed on device.
Through the above description of the embodiments, those skilled in the art can be understood that the present invention can be by
Realized in the mode of software plus required hardware platform, naturally it is also possible to all implemented by software or hardware.Based on this
The understanding of sample, what technical scheme contributed to background technology in whole or in part can be in the form of software product
Embody, the computer software product can be stored in storage medium, such as ROM/RAM, disk, CD, including some
Instruct to cause a computer equipment(Can be personal computer, server, or network equipment etc.)Perform the present invention
Method described in some parts of each embodiment or embodiment.
Each embodiment of the present invention has been described in detail above.However, it should be appreciated by those skilled in the art that do not taking off
In the case of principle and spirit from the present invention, these embodiments can be carried out with various modifications, combination or sub-portfolio, and so
Modification should fall within the scope of the present invention.
Claims (18)
1. a kind of information processing method, applied to one or more mobile electronic equipment, the mobile electronic equipment quilt
Instant positioning and map structuring for circumstances not known, it is characterised in that methods described includes:
N rank, wherein n >=2 are created when the instant positioning of the circumstances not known is with map structuring;
In the characteristic information of the i-stage not repetition that the multiple nodes of middle filtering are included, wherein 1≤i≤n;
It is single node by the multiple node stipulations;
Before the step of characteristic information of the repetition included in the not middle multiple nodes of filtering of i-stage, methods described is also
Including:
When 1<During i≤n, the i-th scene is created;
Obtain the single node of institute's stipulations in the i-th -1 rank, the present node as the i-th scene;
Local map structure is carried out to the present node and history node included in the i-th scene;
It is described by the multiple node stipulations be single node the step of after, methods described also includes:When 1<i<During n,
Retain last predetermined number destination node in i-th scene, to continue i-stage other positioning and map structuring immediately;
Delete remaining node;And
The single node is sent to i+1 rank, to carry out the instant positioning of i+1 rank and map structuring.
2. method according to claim 1, it is characterised in that in the weight included in the not middle multiple nodes of filtering of i-stage
Before the step of multiple characteristic information, methods described also includes:
As i=1, the first sub-scene of the i-th scene is created;
Obtain frame observation data, the present node as the first sub-scene;
Local map structure is carried out to the present node and history node included in the first sub-scene.
3. method according to claim 2, it is characterised in that in the present node to including in the first sub-scene and go through
After history node carries out the step of local map structure, methods described also includes:
When the characteristic information of the present node is enough, and works as the present node to including in the first sub-scene and go through
When history node progress local map successfully constructs, the present node is added as into history node;
Judge whether the total number of the present node and history node included in the first sub-scene is more than first threshold;
If the present node and the total number of history node that include in the first sub-scene are more than first threshold, further sentence
It is disconnected to whether there is the second sub-scene;
If there is no the second sub-scene, then the second sub-scene of the i-th scene is created, and present node is added as into the second subfield
The history node of scape;
If there is the second sub-scene, then present node is directly added as to the history node of the second sub-scene.
4. method according to claim 3, it is characterised in that the repetition included in the not middle multiple nodes of filtering of i-stage
Characteristic information the step of include:
Judge whether the total number of the present node and history node included in the second sub-scene is more than Second Threshold;
If the present node and the total number of history node that include in the second sub-scene are more than Second Threshold, filtering is current
The characteristic information for the repetition that node and history node are included.
5. method according to claim 3, it is characterised in that it is described by the multiple node stipulations be single node the step of
Afterwards, methods described also includes:
It is discarded in all nodes that the first sub-scene includes;
Using the second sub-scene as the first sub-scene, to continue i-stage other positioning and map structuring immediately;
Cancel second sub-scene;And
The single node is sent to i+1 rank, to carry out the instant positioning of i+1 rank and map structuring.
6. method according to claim 1, it is characterised in that in the present node and history to including in the i-th scene
After node carries out the step of local map structure, methods described also includes:
When the characteristic information of the present node is enough, and when the present node and history to including in the i-th scene
When node progress local map successfully constructs, the present node is added as into history node.
7. method according to claim 1, it is characterised in that the repetition included in the not middle multiple nodes of filtering of i-stage
Characteristic information the step of include:
Judge whether the total number of the present node and history node included in the i-th scene is more than the 3rd threshold value;
If the present node and the total number of history node that include in the i-th scene are more than the 3rd threshold value, filtering is worked as
The characteristic information for the repetition that front nodal point and history node are included.
8. method according to claim 1, it is characterised in that it is described by the multiple node stipulations be single node the step of
Afterwards, methods described also includes:
As i=n, the single node is shown.
9. method according to claim 1, it is characterised in that described the step of by the multiple node stipulations being single node wraps
Include:
The characteristic information of the repetition included to multiple nodes assigns weighted value;
It is weighted averagely come the characteristic information of the repetition included to multiple nodes according to the weighted value;
Generate the single node of the characteristic information after including weighted average;And
Delete the multiple node.
10. a kind of information processor, applied to one or more mobile electronic equipment, the mobile electronic equipment quilt
Instant positioning and map structuring for circumstances not known, it is characterised in that described device includes:
Level creation unit, for creating n rank in instant positioning and the map structuring of the circumstances not known, wherein n >=
2;
Characteristic filter unit, for the characteristic information in the i-stage not repetition that the multiple nodes of middle filtering are included, wherein 1≤i≤
n;
Node stipulations unit, for being single node by the multiple node stipulations;
Described device also includes:
Scene creation unit, for when 1<During i≤n, wrapped in the characteristic filter unit in the not middle multiple nodes of filtering of i-stage
Before the characteristic information of the repetition contained, the i-th scene is created;
Node acquiring unit, obtain the single node of institute's stipulations in the i-th -1 rank, the present node as the i-th scene;
Map constructing unit, local map structure is carried out to the present node and history node included in the i-th scene;
Processing unit, for when 1<i<During n, the node stipulations unit by the multiple node stipulations be single node it
Afterwards, last predetermined number destination node is retained in the i-th scene, to continue i-stage other positioning and map structuring immediately;
Delete remaining node;And the single node is sent to i+1 rank, so as to carry out the instant positioning of i+1 rank with
Map structuring.
11. device according to claim 10, it is characterised in that described device also includes:
Scene creation unit, for as i=1, being included in the characteristic filter unit in the not middle multiple nodes of filtering of i-stage
Repetition characteristic information before, create the i-th scene the first sub-scene;
Node acquiring unit, for obtaining frame observation data, the present node as the first sub-scene;
Map constructing unit, local map structure is carried out for the present node to including in the first sub-scene and history node
Build.
12. device according to claim 11, it is characterised in that described device also includes:
Node adding device, in the map constructing unit to the present node and history section that include in the first sub-scene
After point carries out local map structure, when the characteristic information of the present node is enough, and when in the first sub-scene
When present node and history node the progress local map included successfully constructs, the present node is added as into history node,
Whether the present node and the total number of history node that the scene creation unit judges include in the first sub-scene are big
In first threshold, and if the present node and the total number of history node included in the first sub-scene is more than the first threshold
Value, then further determine whether the second sub-scene be present;If there is no the second sub-scene, then the scene creation unit establishment
Second sub-scene of the i-th scene, and notify the node adding device that present node is added as to the history section of the second sub-scene
Point;If there is the second sub-scene, then directly notify the node adding device that present node is added as into going through for the second sub-scene
History node.
13. device according to claim 12, it is characterised in that the characteristic filter unit judges include in the second sub-scene
Present node and the total number of history node whether be more than Second Threshold;If in the present node that the second sub-scene includes
It is more than Second Threshold with the total number of history node, then filters the feature letter for the repetition that present node and history node are included
Breath.
14. device according to claim 12, it is characterised in that described device also includes:
Processing unit, after by the multiple node stipulations being single node in the node stipulations unit, it is discarded in the
All nodes that one sub-scene includes;It is other immediately to continue i-stage using the second sub-scene as the first sub-scene
Positioning and map structuring;Cancel second sub-scene;And the single node is sent to i+1 rank, to carry out
The instant positioning of i+1 rank and map structuring.
15. device according to claim 10, it is characterised in that described device also includes:
Node adding device, in the map constructing unit to the present node and history node that include in the i-th scene
After carrying out local map structure, when the characteristic information of the present node is enough, and when to including in the i-th scene
Present node and history node carry out local map when successfully constructing, the present node is added as into history node.
16. device according to claim 10, it is characterised in that what the characteristic filter unit judges included in the i-th scene
Whether the total number of present node and history node is more than the 3rd threshold value;If in the present node that the i-th scene includes and go through
The total number of history node is more than the 3rd threshold value, then filters the feature letter for the repetition that present node and history node are included
Breath.
17. device according to claim 10, it is characterised in that described device also includes:
Processing unit, for as i=n, after the multiple node stipulations are single node by the node stipulations unit,
Show the single node.
18. device according to claim 10, it is characterised in that the repetition that the node stipulations unit is included to multiple nodes
Characteristic information assign weighted value;It is weighted according to the weighted value come the characteristic information of the repetition included to multiple nodes
It is average;Generate the single node of the characteristic information after including weighted average;And delete the multiple node.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201210309510.8A CN103631261B (en) | 2012-08-27 | 2012-08-27 | Information processing method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201210309510.8A CN103631261B (en) | 2012-08-27 | 2012-08-27 | Information processing method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103631261A CN103631261A (en) | 2014-03-12 |
CN103631261B true CN103631261B (en) | 2017-11-28 |
Family
ID=50212449
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201210309510.8A Active CN103631261B (en) | 2012-08-27 | 2012-08-27 | Information processing method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103631261B (en) |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103632606B (en) * | 2012-08-27 | 2016-03-30 | 联想(北京)有限公司 | Information processing method and device |
CN105527968A (en) * | 2014-09-29 | 2016-04-27 | 联想(北京)有限公司 | Information processing method and information processing device |
US10311629B2 (en) * | 2016-01-22 | 2019-06-04 | Intel Corporation | Level of detail selection during ray tracing |
US20190333239A1 (en) * | 2016-12-02 | 2019-10-31 | Cloudminds (Shenzhen) Robotics Systems Co., Ltd. | Positioning method and device |
US10921816B2 (en) * | 2017-04-21 | 2021-02-16 | Korea Advanced Institute Of Science And Technology | Method and apparatus for producing map based on hierarchical structure using 2D laser scanner |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2011043419A (en) * | 2009-08-21 | 2011-03-03 | Sony Corp | Information processor, information processing method, and program |
CA2812723C (en) * | 2010-09-24 | 2017-02-14 | Evolution Robotics, Inc. | Systems and methods for vslam optimization |
KR101772977B1 (en) * | 2010-10-07 | 2017-08-31 | 삼성전자주식회사 | Moving robot and map-building method thereof |
CN103632606B (en) * | 2012-08-27 | 2016-03-30 | 联想(北京)有限公司 | Information processing method and device |
-
2012
- 2012-08-27 CN CN201210309510.8A patent/CN103631261B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN103631261A (en) | 2014-03-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Sadeghian et al. | Sophie: An attentive gan for predicting paths compliant to social and physical constraints | |
KR102296507B1 (en) | Method for tracking object by using convolutional neural network including tracking network and computing device using the same | |
CN103632606B (en) | Information processing method and device | |
CN103631261B (en) | Information processing method and device | |
Zhou et al. | A deep Q-network (DQN) based path planning method for mobile robots | |
Van Den Berg et al. | Interactive navigation of multiple agents in crowded environments | |
CN110523081A (en) | The method and device for planning in navigation pathfinding path | |
Kallmann et al. | Geometric and discrete path planning for interactive virtual worlds | |
JP2018109621A (en) | Navigation for moving body | |
CN111260072A (en) | Reinforced learning exploration method based on generation countermeasure network | |
Cui et al. | Direction oriented pathfinding in video games | |
CN110132282A (en) | Unmanned plane paths planning method and device | |
Barnett et al. | Coordinated crowd simulation with topological scene analysis | |
WO2012069022A1 (en) | Method for animating characters, with collision avoidance based on tracing information | |
WO2020181934A1 (en) | Method and device for determining position of target object on the basis of particle swarm algorithm | |
Wang et al. | An immersive multi-agent system for interactive applications | |
CN105701573B (en) | A kind of route processing method and device based on GIS buffer zone analysis | |
Zhao et al. | Hybrid navigation method for multiple robots facing dynamic obstacles | |
WO2021008798A1 (en) | Training of a convolutional neural network | |
Geraerts et al. | Enhancing corridor maps for real-time path planning in virtual environments | |
CN110490803A (en) | A kind of joining method, equipment and the storage medium of image, semantic segmentation block prediction | |
Pan et al. | Learning navigation policies for mobile robots in deep reinforcement learning with random network distillation | |
CN113741461B (en) | Multi-robot obstacle avoidance method oriented to limited communication under complex scene | |
Sudkhot et al. | Collision avoidance and path planning in crowd simulation | |
CN113432610A (en) | Robot passage planning method and device, robot and storage medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |