CN109471446A - A kind of complete coverage path planning method neural network based - Google Patents
A kind of complete coverage path planning method neural network based Download PDFInfo
- Publication number
- CN109471446A CN109471446A CN201811325956.3A CN201811325956A CN109471446A CN 109471446 A CN109471446 A CN 109471446A CN 201811325956 A CN201811325956 A CN 201811325956A CN 109471446 A CN109471446 A CN 109471446A
- Authority
- CN
- China
- Prior art keywords
- auv
- path
- barrier
- neuron
- neural network
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course or altitude of land, water, air, or space vehicles, e.g. automatic pilot
- G05D1/04—Control of altitude or depth
- G05D1/06—Rate of change of altitude or depth
- G05D1/0692—Rate of change of altitude or depth specially adapted for under-water vehicles
Abstract
The invention discloses a kind of complete coverage path planning methods neural network based, comprising: utilizes the environmental information around preposition sonar model detection AUV, the three-dimensional underwater grating map of building；Environmental information is mapped to biology to inspire on neural network model, and is introducing excitation input signal in the corresponding neuron of grid with not covering for grating map, calculates the active output valve of each neuron；All standing path is planned according to the detection mission that the active output valve of each neuron and direction-guidance function are AUV；After AUV enters dead zone, dead zone is quickly fled from using network reset method；If meeting irregular barrier, optimize all standing path using agent approach；Finally judge whether to complete all standing to detection target, if completed, terminates detection mission；Otherwise, it repeats the above steps.The present invention effectively reduces repetition covering problem, reaches path optimization's effect in such a way that biology inspires neural network and templating path planning to combine.
Description
Technical field
The present invention relates to underwater robot Navigation Control fields, and in particular to a kind of all standing path neural network based
Planing method.
Background technique
Path planning (the Path of autonomous underwater robot AUV (Autonomous Underwater Vehicle)
Planning), refer to that robot perceives underwater environment using known environment information or according to self-sensor device, voluntarily
Cook up one efficiently, safety, without the navigation route that touches, to support the exploration to submarine target region, or execute related make
Industry mission；Visible path planning is one of core research contents of AUV technology.But since the working environment of benthos is often
Badly, complicated and it is difficult to predict, while the limitation of AUV itself perception, therefore the underwater path planning of AUV is than ground moving machine
The more complicated difficulty of the path planning of device people and airborne aircraft, more challenge.
For AUV, since underwater environment is complicated, changeable, it is difficult to predict research achievement is still less, and research is main
It concentrates in template matching, map structuring and the research of artificial intelligence path planning；The Path selection path template matching AUV is advised
It draws, the either simple case paths planning method of early stage, or the adaptive Sample Method studied recently, it is dynamic all to there is environment
State change when, Path selection template matching fail the problem of；Although map structuring can be fitted by online updating raster data
The dynamic change of underwater environment is answered, but frequently due to underwater environment interference, while AUV sensor resource is limited, makes under water
The integrity problem of figure building is not solved effectively always, and the failure of Path selection template matching, path planning is caused to repeat to cover
Lid.
Summary of the invention
The technical problem to be solved by the present invention is to existing route planning technologies there are the failure of Path selection template matching, again
The problem of covering again.
In order to solve the above-mentioned technical problem, the technical scheme adopted by the invention is that providing a kind of neural network based complete
Overlay path planing method, comprising the following steps:
Step S10, using the environmental information around preposition sonar model detection AUV, the three-dimensional underwater grid of dynamic construction
Figure；
Step S20, by the environmental information detected be mapped to biology inspire neural network model on, and with grating map
Do not cover in the corresponding neuron of grid introduce excitation input signal, calculate the active output valve of each neuron；
Step S30, the active output valve and direction-guidance letter of each neuron of neural network model are inspired according to biology
The detection mission that number is AUV plans all standing path；
Step S40, judgement works as whether AUV enters dead zone, if so, executing step S50；Otherwise, step S60 is executed；
Step S50, dead zone is quickly fled from using network reset method；
Step S60, judge whether to meet irregular barrier, if so, executing step S70；Otherwise, step is executed
S80；
Step S70, optimize all standing path using agent approach；
Step S80, judge whether to complete all standing to detection target, if completed, terminate detection mission；Otherwise, it executes
Step S10.
In the above-mentioned methods, the environmental information that sonar acquires is converted into grating map each by the preposition sonar model
The brief inference value of grid has obtained the brief inference value of each grid, and according to specified trellis states could decision rule, judgement
The state of each grid out, the underwater grating map of dynamic construction.
In the above-mentioned methods, according to the size of the distance between tested barrier and AUV r and acoustic irradiation radius range R
Relationship needs the belief function of two different sonars to carry out the calculating of belief function apportioning cost；
As R-d≤r≤R+d, which is referred to as first interval I, the belief function apportioning cost in the section are as follows:
Work as RminWhen≤r≤R-d, which is referred to as second interval II, the belief function apportioning cost in the section are as follows:
Wherein, mO(i, j, k) is the belief function apportioning cost that grid occupies barrier；mE(i, j, k) is the non-barrier of grid
Belief function apportioning cost；m{O、E}(i, j, k) is the belief function apportioning cost of grid nondeterministic statement mode；I, j, k indicate grid
The coordinate value of horizontal axis, the longitudinal axis, vertical pivot in lattice map；O indicates barrier；E indicates non-barrier；D is radiation error；α is radiation
Angle；The search angle of sonar is β, calculation formula are as follows:
x′e、y′e、z′eRespectively indicate the coordinate value of barrier horizontal axis, longitudinal axis and vertical pivot in inertial coodinate system；xr、yr、zr
Respectively indicate the coordinate value of AUV horizontal axis, longitudinal axis and vertical pivot in inertial coodinate system；A=(x'e-xr, 0,0) and it is known in sonar biography
Certain point coordinate on the central axis of sensor, B=(x'e-xr,y'e-yr,z'e-zr) be a certain barrier inertial coordinate, thenIndicate AUV to barrier vector,Indicate the vector of certain point A on AUV to central axis.
In the above-mentioned methods, step S20 specifically:
According to grating map and biology inspire neural network model mapping relations, to biology inspire neural network model into
Row Initialize installation；
By on grating map, uncovered area will be set to by the AUV zone state for executing region overlay
" Uncovered ", and inspire the neuron of neural network model corresponding position to introduce pumping signal+E in biology；When AUV is executed
After region overlay, corresponding region state is changed to overlay area " Covered ", and by the pumping signal of corresponding position neuron
Zero setting；
On grating map, barrier region state is set to barrier region " Occupied ", and open in biology
Go crazy network model corresponding position neuron introduce inhibit input signal-E；
The active output valve of each neuron is finally calculated according to the input signal of each neuron.
In the above-mentioned methods, step S30 specifically:
Based on bow-shaped route walking manner, horizontal guidance is added in routing strategy and carries out path optimization, AUV
The calculation formula of the path P ath of selection are as follows:
Wherein, PnIndicate the next step position in the path AUV on grating map；Indicate all next step positions (nerve
Member) in maximum neuron activity value；xklIndicate the activity value of the periphery neuron k adjacent neurons；C is the normal of a positive value
Number；ylIt is with AUV bow to deflection angle Δ ψlA relevant monotonic function, is defined as follows:
Wherein, Pc、PpAnd PnIt is illustrated respectively in the current location in the path AUV on map, back position and next step
It sets,WithThe current location of AUV, back position and next step position respectively on map
Coordinate.
In the above-mentioned methods, quickly to flee from dead zone using network reset method specific as follows:
The positive activity value of the neuron of overlay area is reset, direction coefficient c=0 is enabled；
Neural network model recalculates and generates new active Distribution value；
After AUV, which leaves dead zone, reaches uncovering area, restore c initial value.
It in the above-mentioned methods, include four Path selection templates in agent approach, when meeting irregular barrier in step S70
When hindering object, by matching different Path selection templates completions to the path optimization at the edge of irregular slalom object；
8, the current location AUV periphery position is carried out numbering clockwise from No. 1 to 8 since the Angle Position of lower-left, four roads
Diameter selection template is specifically defined are as follows:
First path selects template, if AUV is moved from the bottom up, and there is barrier in position immediately ahead of AUV, judges AUV
No. 6 location status whether be not cover；
If No. 6 location status are not cover, No. 6 positions of next step Path selection of AUV, and judge No. 7 position shapes
Whether state is not cover；If No. 7 location status are not cover, No. 7 positions of next step Path selection of AUV No. 6 positions；
Second Path selection template, if there is barrier in the lower left of AUV, and AUV is moved from the bottom up, judges the 2 of AUV
Whether number location status is not cover；
If No. 2 location status of AUV are not cover, No. 2 positions of the Path selection AUV of next step；
Third Path selection template, if there is barrier on the upper left side of AUV, and AUV is moved from top to bottom, judges the 2 of AUV
Whether number location status is not cover；
If No. 2 location status of AUV are not cover, No. 2 positions of the Path selection AUV of next step；
4th Path selection template, if there is barrier in the lower right of AUV, and AUV is moved from right to left, judges the 8 of AUV
Whether number location status is not cover；
If No. 8 location status of AUV are not cover, No. 8 positions of the Path selection AUV of next step.
In the above-mentioned methods, if corresponding template can not be matched in four kinds of Path selection templates, according to step S30
Method carry out next step Path selection.
Compared with prior art, the invention has the following advantages that
1, the distance that can be distinguished in water using preposition sonar sensor is remote, and resolution ratio is higher, while handling the mistake of information
Journey is simple, strong real-time and cheap；
2, in such a way that biology inspires neural network and templating path planning to combine, repetition covering is effectively reduced
Problem reaches path optimization's effect.
Detailed description of the invention
Fig. 1 is the flow chart of kind of complete coverage path planning method neural network based of the invention；
Fig. 2 is Forward-looking Sonar model in the present invention；
Fig. 3 is the schematic diagram that Glasius biology inspires neural network model in the present invention；
Fig. 4 is direction-guidance schematic diagram in the present invention；
Fig. 5 is that first path selects template path to select schematic diagram in the present invention；
Fig. 6 is that the second Path selection template path selects schematic diagram in the present invention；
Fig. 7 is that third Path selection template path selects schematic diagram in the present invention；
Fig. 8 is that the 4th Path selection template path selects schematic diagram in the present invention.
Specific embodiment
The present invention executes the task of underwater regional area all standing search for AUV, realizes the path rule of map all standing
It draws, using the neural dynamics DYNAMIC DISTRIBUTION feature of neural network, based on being traversed by bow-shaped route, in conjunction with the fortune of robot
Dynamic direction, formulates the strategy of Path selection and region overlay, and to the edge of irregular slalom object, formulates limited Path selection
Template not only avoids the occurrence of the failure of Path selection template matching as supplement, but also effectively reduces repetition covering problem, reaches
Path optimization's effect.The present invention is described in detail with specific embodiment with reference to the accompanying drawings of the specification.
As shown in Figure 1, a kind of complete coverage path planning method neural network based, comprising the following steps:
Step S10, using the environmental information around preposition sonar model detection AUV, the three-dimensional underwater grid of dynamic construction
Figure.
In the present invention, when the underwater map of dynamic construction three-dimensional, the present invention is right using preposition sonar model (as shown in Figure 2)
Environmental information around AUV is acquired, and the environmental information that sonar acquires accurately can be converted into grid by preposition sonar model
The brief inference value of each grid in map, after having obtained the brief inference value of each grid, so that it may according to specified grid
Condition discrimination rule judges the state of each grid, to achieve the purpose that dynamic construction map.For a three-dimensional environment
Used in sonar model because of its symmetry the detection model of the sonar need to be only discussed on a two-dimensional surface.Such as Fig. 2
It is shown, if acoustic irradiation radius range is R, radiation error d, angle of radiation α, it is tested the distance between barrier and AUV
Search angle for r, sonar is β, and the inertial coordinate for the correspondence barrier being converted to according to front coordinate acquires, such as a certain obstacle
The inertial coordinate of object is B=(x'e-xr,y'e-yr,z'e-zr), wherein x'e、y'e、z'eBarrier is respectively indicated in inertial coordinate
The coordinate value of horizontal axis, the longitudinal axis and vertical pivot in system；xr、yr、zrRespectively indicate AUV horizontal axis, the longitudinal axis and vertical pivot in inertial coodinate system
Coordinate value；Correspondingly, the known certain point A=(x' on the central axis of sonar sensore-xr, 0,0), it can then acquire:
Wherein,Indicate AUV to barrier vector,Indicate the vector of certain point A on AUV to central axis；
According to the size relation that the distance between tested barrier and AUV are r and acoustic irradiation radius range is R, need
The calculating of belief function apportioning cost is carried out using the belief function of two different sonars；
As R-d≤r≤R+d, which is referred to as first interval I, the belief function apportioning cost in the section are as follows:
Work as RminWhen≤r≤R-d, which is referred to as second interval II, the belief function apportioning cost in the section are as follows:
Wherein, mO(i, j, k) is the belief function apportioning cost that grid occupies barrier；mE(i, j, k) is the non-barrier of grid
Belief function apportioning cost；m{O、E}(i, j, k) is the belief function apportioning cost of grid nondeterministic statement mode；I, j, k indicate grid
The coordinate value of horizontal axis, the longitudinal axis, vertical pivot in lattice map；O indicates barrier；E indicates non-barrier.
Step S20, the environmental information around the AUV detected biology is mapped to inspire on neural network model, and
Excitation input signal is introduced in the corresponding neuron of grid with not covering for grating map, the activity for calculating each neuron is defeated
It is worth out.
Present invention employs Glasius biologies to inspire neural network model (as shown in Figure 3).With the two-dimentional ring of search system
For the structure of border, each neuron in Glasius biology inspiration neural network model is mapped with two-dimensional grating map,
A position of the autonomous underwater robot in grating map is represented with each neuron；In this model, excitation input
From the neuron (uncovered region grid) for representing target, and known excitation is only derived from the nerve for representing barrier
Member goes out the traveling-position of autonomous underwater robot by the distribution situation of neuronal activity output valve come decision.In the present invention
Step S20 specifically:
According to grating map and biology inspire neural network model mapping relations, to biology inspire neural network model into
Row Initialize installation；
By on grating map, uncovered area will be set to by the AUV zone state for executing region overlay
" Uncovered ", and inspire the neuron of neural network model corresponding position to introduce pumping signal+E in biology；When AUV is executed
After region overlay, corresponding region state is changed to overlay area " Covered ", and by the pumping signal of corresponding position neuron
Zero setting；
On grating map, barrier region state is set to barrier region " Occupied ", and open in biology
Go crazy network model corresponding position neuron introduce inhibit input signal-E；
As it can be seen that biology inspires the corresponding relationship between the neuron external input signal and map state of neural network model
Are as follows:
The active output valve of each neuron is finally calculated according to the input signal of each neuron.
Step S30, the active output valve and direction-guidance letter of each neuron of neural network model are inspired according to biology
The detection mission that number is AUV plans all standing path；Specific path planning is as follows:
Based on bow-shaped route walking manner, horizontal guidance is added in routing strategy and carries out path optimization, AUV
The calculation formula of the path P ath of selection are as follows:
Wherein, PnIndicate the next step position in the path AUV on grating map；Indicate all next step positions (nerve
Member) in maximum neuron activity value；xklIndicate the activity value of the periphery neuron k adjacent neurons；C is the normal of a positive value
Number；ylIt is with AUV bow to deflection angle Δ ψlA relevant monotonic function, is defined as follows:
Wherein, Pc、PpAnd PnIt is illustrated respectively in the current location in the path AUV on map, back position and next step
It sets,WithThe current location of AUV, back position and next step position respectively on map
Coordinate.
It is illustrated in figure 4 the calculating of direction-guidance schematic diagram in the present invention, when AUV next step path keeps working as front direction not
Become, then AUV bow is to deflection angle Δ ψl=0；When AUV next step position is to shrink back, then AUV bow is to deflection angle Δ ψl=π,
Δψl∈[0,π]；Then according to the calculation formula of the path P ath of AUV selection it is found that (direction is not when AUV straight forward
Become), yl=1；When AUV next step position is to shrink back, yl=0；When AUV next step position and present bit are equipped with certain angle
It turns to, 0 < yl< 1；So AUV selection path P ath calculation formula can be understood as AUV selection path next step when
It waits, while considering next step position and corresponding to the neural dynamics size and steering of neuron, so that cooking up the road come
Diameter keeps keeping straight on as far as possible, reduces and turns to, and avoids repeating to cover.
Step S40, judgement works as whether AUV enters dead zone, if so, executing step S50；Otherwise, step S60 is executed.
Step S50, dead zone is quickly fled from using network reset method；Network reset method is specific as follows:
The positive activity value of the neuron of overlay area is reset, enabling direction coefficient c=0, (namely cancellation horizontal guidance, changes
With point-to-point paths planning method), it allows neural network model to recalculate and generates new active Distribution value, when AUV leaves
After deadlock area reaches uncovering area, restore c initial value.
Step S60, judge whether to meet irregular barrier, if so, executing step S70；Otherwise, step is executed
S80。
Step S70, when meeting irregular barrier, optimize all standing path using agent approach.
In the present invention, complete coverage path planning does not require nothing more than avoiding obstacles, also goes access irregular using agent approach
The edge and every nook and cranny of barrier.Four Path selection templates are formulated in agent approach of the invention to complete irregularly to hinder
The path optimization for hindering object edge, (wherein, C indicates the current position AUV, and P indicates the path AUV back as shown in Fig. 5 to Fig. 8
Position, obstacles indicate barrier, 8, the current location AUV periphery position is carried out since the Angle Position of lower-left from No. 1
Numbered clockwise to 8), specifically:
First path selects template.
The template definition is that AUV is moved from the bottom up, and there are barrier and No. 6 position shapes of AUV in position immediately ahead of AUV
State is next step routing strategy when not covering, and preferential selection is turned right, to carry out direction gage when turning round to AUV
Model guarantees the turn for doing right angle detour as far as possible.As shown in figure 5, first path template definition is as follows:
If (having barrier immediately ahead of AUV) then
If (No. 6 positions do not cover) then
{ path selects No. 6 positions in next step }
If (No. 7 positions do not cover) then
{ path selects No. 7 positions in next step }
end if
end if
end if
I.e. if AUV is moved from the bottom up, and there is barrier in position immediately ahead of AUV, judges that No. 6 location status of AUV are
No is not cover；
If No. 6 location status are not cover, No. 6 positions of next step Path selection of AUV, and judge No. 7 position shapes
Whether state is not cover；If No. 7 location status are not cover, No. 7 positions of next step Path selection of AUV No. 6 positions；
Otherwise, the Path selection of next step is carried out according to the method for step S30.
Second Path selection template.
The template definition is that AUV is moved from the bottom up, and the lower left AUV position is with the presence of barrier and the left side AUV is not visited
Next step routing strategy when the region asked, the region in preferential access left side, selects next step path to turn left, such as at this time
Shown in Fig. 6, the second path template definition is as follows:
If (lower left that barrier is located at AUV) then
If (AVU is from the bottom up) then
If (No. 2 positions do not cover) then
{ path selects No. 2 positions in next step }
end if
end if
end if
I.e. if there is barrier in the lower left of AUV, and AUV is moved from the bottom up, judge AUV No. 2 location status whether
Not cover；
If No. 2 location status of AUV are not cover, No. 2 positions of the Path selection AUV of next step；
Otherwise, it is not moved according to template, the Path selection of next step is carried out according to the method for step S30.
Third Path selection template.
The template definition is that AUV is moved from top to bottom, and the upper left side AUV position is with the presence of barrier and the left side AUV is not visited
Next step routing strategy when the region asked, the region in preferential access left side, selects next step path to turn left, such as at this time
Shown in Fig. 7, third path template definition is as follows:
If (upper left side that barrier is located at AUV) then
If (AUV is moved from top to bottom) then
If (No. 2 positions do not cover) then
{ path selects No. 2 positions in next step }
end if
end if
end if
I.e. if there is barrier on the upper left side of AUV, and AUV is moved from top to bottom, judge AUV No. 2 location status whether
Not cover；
If No. 2 location status of AUV are not cover, No. 2 positions of the Path selection AUV of next step；
Otherwise, it is not moved according to template, the Path selection of next step is carried out according to the method for step S30.
4th Path selection template.
The template definition is that AUV is moved from right to left, AUV lower right position with the presence of barrier and immediately below AUV not
Next step routing strategy when the region of access, preferentially accesses the region at this time, as shown in figure 8, third path template is fixed
Justice is as follows:
If (barrier is in the lower right of AUV) then
If (AUV is moved to the left) then
If (No. 8 positions do not cover) then
{ path selects No. 8 positions in next step }
end if
end if
end if
I.e. if there is barrier in the lower right of AUV, and AUV is moved from right to left, judge AUV No. 8 location status whether
Not cover；
If No. 8 location status of AUV are not cover, No. 8 positions of the Path selection AUV of next step；
Otherwise, it is not moved according to template, the Path selection of next step is carried out according to the method for step S30.
Step S80, judge whether to complete all standing to detection target, if completed, AUV returns to designated position, terminates
Detection mission；Otherwise, step S10 is executed, detection mission is continued.
The invention is not limited to above-mentioned preferred forms, and anyone should learn that is made under the inspiration of the present invention
Structure change, the technical schemes that are same or similar to the present invention are fallen within the scope of protection of the present invention.
Claims (8)
1. a kind of complete coverage path planning method neural network based, which comprises the following steps:
Step S10, the environmental information around preposition sonar model detection AUV, the three-dimensional underwater grating map of dynamic construction are utilized；
Step S20, by the environmental information detected be mapped to biology inspire neural network model on, and with grating map not
It covers and introduces excitation input signal in the corresponding neuron of grid, calculate the active output valve of each neuron；
Step S30, it is according to the active output valve of each neuron of biology inspiration neural network model and direction-guidance function
The detection mission of AUV plans all standing path；
Step S40, judge whether current AUV enters dead zone, if so, executing step S50；Otherwise, step S60 is executed；
Step S50, dead zone is quickly fled from using network reset method；
Step S60, judge whether to meet irregular barrier, if so, executing step S70；Otherwise, step S80 is executed；
Step S70, optimize all standing path using agent approach；
Step S80, judge whether to complete all standing to detection target, if completed, terminate detection mission；Otherwise, step is executed
S10。
2. the method according to claim 1, wherein the environmental information that the preposition sonar model acquires sonar
It is converted into the brief inference value of each grid in grating map, has obtained the brief inference value of each grid, and according to specified
Trellis states could decision rule judges the state of each grid, the underwater grating map of dynamic construction.
3. according to the method described in claim 2, it is characterized in that, according to the distance between tested barrier and AUV r and sound wave
The size relation of radiation radius range R needs the belief function of two different sonars to carry out the calculating of belief function apportioning cost；
As R-d≤r≤R+d, which is referred to as first interval I, the belief function apportioning cost in the section are as follows:
Work as RminWhen≤r≤R-d, which is referred to as second interval II, the belief function apportioning cost in the section are as follows:
Wherein, mO(i, j, k) is the belief function apportioning cost that grid occupies barrier；mE(i, j, k) is the letter of the non-barrier of grid
Spend function apportioning cost；m{O、E}(i, j, k) is the belief function apportioning cost of grid nondeterministic statement mode；I, j, k indicate grid
The coordinate value of horizontal axis, the longitudinal axis, vertical pivot in figure；O indicates barrier；E indicates non-barrier；D is radiation error；α is angle of radiation；
The search angle of sonar is β, calculation formula are as follows:
x′e、y′e、z′eRespectively indicate the coordinate value of barrier horizontal axis, longitudinal axis and vertical pivot in inertial coodinate system；xr、yr、zrRespectively
Indicate the coordinate value of AUV horizontal axis, longitudinal axis and vertical pivot in inertial coodinate system；A=(x'e-xr, 0,0) and it is known in sonar sensor
Central axis on certain point coordinate, B=(x'e-xr,y'e-yr,z'e-zr) be a certain barrier inertial coordinate, then
Indicate AUV to barrier vector,Indicate the vector of certain point A on AUV to central axis.
4. the method according to claim 1, wherein step S20 specifically:
The mapping relations that neural network model is inspired according to grating map and biology inspire neural network model to carry out just biology
Beginningization setting；
By on grating map, uncovered area " Uncovered " will be set to by the AUV zone state for executing region overlay,
And the neuron of neural network model corresponding position is inspired to introduce pumping signal+E in biology；It, will after AUV executes region overlay
Corresponding region state is changed to overlay area " Covered ", and by the pumping signal zero setting of corresponding position neuron；
On grating map, barrier region state is set to barrier region " Occupied ", and inspires mind in biology
Neuron through network model corresponding position, which introduces, inhibits input signal-E；
The active output valve of each neuron is finally calculated according to the input signal of each neuron.
5. the method according to claim 1, wherein step S30 specifically:
Based on bow-shaped route walking manner, horizontal guidance is added in routing strategy and carries out path optimization, AUV selection
Path P ath calculation formula are as follows:
Wherein, PnIndicate the next step position in the path AUV on grating map；It indicates in all next step positions (neuron)
The activity value of maximum neuron；xklIndicate the activity value of the periphery neuron k adjacent neurons；C is the constant of a positive value；yl
It is with AUV bow to deflection angle Δ ψlA relevant monotonic function, is defined as follows:
Wherein, Pc、PpAnd PnIt is illustrated respectively in the current location in the path AUV on map, back position and next step position,
(xPc,yPc)、(xPp,yPp) and (xPl,yPl) it is respectively the current location of AUV on map, back position and next step position
Coordinate.
6. the method according to claim 1, wherein in step s 50, quickly being fled from extremely using network reset method
Area is specific as follows:
The positive activity value of the neuron of overlay area is reset, direction coefficient c=0 is enabled；
Neural network model recalculates and generates new active Distribution value；
After AUV, which leaves dead zone, reaches uncovering area, restore c initial value.
7. the method according to claim 1, wherein including four Path selections in agent approach in step S70
Template is completed by matching different Path selection templates to the edge of irregular slalom object when meeting irregular slalom object
Path optimization；
8, the current location AUV periphery position is carried out numbering clockwise from No. 1 to 8 since the Angle Position of lower-left, four paths choosings
Template is selected to be specifically defined are as follows:
First path selects template, if AUV is moved from the bottom up, and there is barrier in position immediately ahead of AUV, judges No. 6 of AUV
Whether location status is not cover；
If No. 6 location status are not cover, No. 6 positions of next step Path selection of AUV, and judge that No. 7 location status are
No is not cover；If No. 7 location status are not cover, No. 7 positions of next step Path selection of AUV No. 6 positions；
Second Path selection template, if there is barrier in the lower left of AUV, and AUV is moved from the bottom up, judges No. 2 positions of AUV
Set whether state is not cover；
If No. 2 location status of AUV are not cover, No. 2 positions of the Path selection AUV of next step；
Third Path selection template, if there is barrier on the upper left side of AUV, and AUV is moved from top to bottom, judges No. 2 positions of AUV
Set whether state is not cover；
If No. 2 location status of AUV are not cover, No. 2 positions of the Path selection AUV of next step；
4th Path selection template, if there is barrier in the lower right of AUV, and AUV is moved from right to left, judges No. 8 positions of AUV
Set whether state is not cover；
If No. 8 location status of AUV are not cover, No. 8 positions of the Path selection AUV of next step.
8. the method according to claim 1, wherein in step S70, if in four kinds of Path selection templates
It can not be matched to corresponding template, then carry out the Path selection of next step according to the method for step S30.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811325956.3A CN109471446A (en) | 2018-11-08 | 2018-11-08 | A kind of complete coverage path planning method neural network based |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811325956.3A CN109471446A (en) | 2018-11-08 | 2018-11-08 | A kind of complete coverage path planning method neural network based |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109471446A true CN109471446A (en) | 2019-03-15 |
Family
ID=65672409
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811325956.3A Pending CN109471446A (en) | 2018-11-08 | 2018-11-08 | A kind of complete coverage path planning method neural network based |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109471446A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109782779A (en) * | 2019-03-19 | 2019-05-21 | 电子科技大学 | AUV paths planning method under ocean current environment based on population meta-heuristic algorithms |
| CN110989638A (en) * | 2019-12-06 | 2020-04-10 | 南京邮电大学 | Underwater building defect detection method based on autonomous navigation technology |
| CN111158401A (en) * | 2020-01-20 | 2020-05-15 | 北京理工大学 | Distributed unmanned aerial vehicle path planning system and method for encouraging space-time data exploration |
| CN111176281A (en) * | 2019-12-31 | 2020-05-19 | 大连民族大学 | Multi-surface unmanned ship coverage type collaborative search method and system based on quadrant method |
| CN111290398A (en) * | 2020-03-13 | 2020-06-16 | 东南大学 | Unmanned ship path planning method based on biological heuristic neural network and reinforcement learning |
| CN111337930A (en) * | 2020-03-19 | 2020-06-26 | 哈尔滨工程大学 | AUV target searching method |
| CN112352530A (en) * | 2020-10-27 | 2021-02-12 | 点亮明天科技(北京)有限责任公司 | Working path optimization method of automatic weeding robot |
| CN113627646A (en) * | 2021-07-08 | 2021-11-09 | 中汽创智科技有限公司 | Path planning method, device, equipment and medium based on neural network |
| CN114326698A (en) * | 2021-11-17 | 2022-04-12 | 中国船舶重工集团公司第七0九研究所 | UUV coverage detection underwater target task planning method and system |
| CN115143970A (en) * | 2022-09-01 | 2022-10-04 | 安徽大学 | Obstacle avoidance method and system of underwater vehicle based on threat degree evaluation |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH11249734A (en) * | 1998-03-03 | 1999-09-17 | Mitsubishi Heavy Ind Ltd | Autonomous guidance device |
| CN103869824A (en) * | 2014-03-05 | 2014-06-18 | 河海大学常州校区 | Biological antenna model-based multi-robot underwater target searching method and device |
| CN106843216A (en) * | 2017-02-15 | 2017-06-13 | 北京大学深圳研究生院 | A kind of complete traverse path planing method of biological excitation robot based on backtracking search |
| CN108444481A (en) * | 2018-03-25 | 2018-08-24 | 哈尔滨工程大学 | A kind of underwater hiding-machine paths planning method based on aid decision-making system |
| CN108762281A (en) * | 2018-06-08 | 2018-11-06 | 哈尔滨工程大学 | It is a kind of that intelligent robot decision-making technique under the embedded Real-time Water of intensified learning is associated with based on memory |
-
2018
- 2018-11-08 CN CN201811325956.3A patent/CN109471446A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH11249734A (en) * | 1998-03-03 | 1999-09-17 | Mitsubishi Heavy Ind Ltd | Autonomous guidance device |
| CN103869824A (en) * | 2014-03-05 | 2014-06-18 | 河海大学常州校区 | Biological antenna model-based multi-robot underwater target searching method and device |
| CN106843216A (en) * | 2017-02-15 | 2017-06-13 | 北京大学深圳研究生院 | A kind of complete traverse path planing method of biological excitation robot based on backtracking search |
| CN108444481A (en) * | 2018-03-25 | 2018-08-24 | 哈尔滨工程大学 | A kind of underwater hiding-machine paths planning method based on aid decision-making system |
| CN108762281A (en) * | 2018-06-08 | 2018-11-06 | 哈尔滨工程大学 | It is a kind of that intelligent robot decision-making technique under the embedded Real-time Water of intensified learning is associated with based on memory |
Non-Patent Citations (3)
| Title |
|---|
| MINGZHONG YAN: "Complete coverage path planning in an unknown underwater environment based on D-S data fusion real-time map building", 《INTERNATIONAL JOURNAL OF DISTRIBUTED SENSOR NETWORKS》 * |
| 张志远等: "基于生物激励神经网络的清洁机器人遍历路径规划算法的改进", 《轻工学报》 * |
| 曹翔等: "栅格地图中多机器人协作搜索目标", 《控制理论与应用》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109782779A (en) * | 2019-03-19 | 2019-05-21 | 电子科技大学 | AUV paths planning method under ocean current environment based on population meta-heuristic algorithms |
| CN109782779B (en) * | 2019-03-19 | 2020-11-06 | 电子科技大学 | AUV path planning method in ocean current environment based on population hyperheuristic algorithm |
| CN110989638A (en) * | 2019-12-06 | 2020-04-10 | 南京邮电大学 | Underwater building defect detection method based on autonomous navigation technology |
| CN111176281A (en) * | 2019-12-31 | 2020-05-19 | 大连民族大学 | Multi-surface unmanned ship coverage type collaborative search method and system based on quadrant method |
| CN111158401A (en) * | 2020-01-20 | 2020-05-15 | 北京理工大学 | Distributed unmanned aerial vehicle path planning system and method for encouraging space-time data exploration |
| CN111158401B (en) * | 2020-01-20 | 2021-08-27 | 北京理工大学 | Distributed unmanned aerial vehicle path planning system and method for encouraging space-time data exploration |
| CN111290398A (en) * | 2020-03-13 | 2020-06-16 | 东南大学 | Unmanned ship path planning method based on biological heuristic neural network and reinforcement learning |
| CN111290398B (en) * | 2020-03-13 | 2022-10-25 | 东南大学 | Unmanned ship path planning method based on biological heuristic neural network and reinforcement learning |
| CN111337930A (en) * | 2020-03-19 | 2020-06-26 | 哈尔滨工程大学 | AUV target searching method |
| CN111337930B (en) * | 2020-03-19 | 2022-07-15 | 哈尔滨工程大学 | AUV target searching method |
| CN112352530A (en) * | 2020-10-27 | 2021-02-12 | 点亮明天科技(北京)有限责任公司 | Working path optimization method of automatic weeding robot |
| CN113627646A (en) * | 2021-07-08 | 2021-11-09 | 中汽创智科技有限公司 | Path planning method, device, equipment and medium based on neural network |
| CN114326698A (en) * | 2021-11-17 | 2022-04-12 | 中国船舶重工集团公司第七0九研究所 | UUV coverage detection underwater target task planning method and system |
| CN115143970A (en) * | 2022-09-01 | 2022-10-04 | 安徽大学 | Obstacle avoidance method and system of underwater vehicle based on threat degree evaluation |
| CN115143970B (en) * | 2022-09-01 | 2022-11-29 | 安徽大学 | Obstacle avoidance method and system of underwater vehicle based on threat degree evaluation |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109471446A (en) | A kind of complete coverage path planning method neural network based | |
| CN107037809A (en) | A kind of unmanned boat collision prevention method based on improvement ant group algorithm | |
| CN110333739A (en) | A kind of AUV conduct programming and method of controlling operation based on intensified learning | |
| Benjamin et al. | Navigation of unmanned marine vehicles in accordance with the rules of the road | |
| CN108983781A (en) | A kind of environment detection method in unmanned vehicle target acquisition system | |
| Hernández et al. | Online path planning for autonomous underwater vehicles in unknown environments | |
| Hernández et al. | Online motion planning for unexplored underwater environments using autonomous underwater vehicles | |
| KR101831264B1 (en) | Autonomous navigation system and method for a maneuverable platform | |
| Buchanan et al. | Perceptive whole‐body planning for multilegged robots in confined spaces | |
| CN108762264A (en) | The dynamic obstacle avoidance method of robot based on Artificial Potential Field and rolling window | |
| CN109784201B (en) | AUV dynamic obstacle avoidance method based on four-dimensional risk assessment | |
| CN109964188A (en) | Control the method and system of vehicle | |
| CN106873599A (en) | Unmanned bicycle paths planning method based on ant group algorithm and polar coordinate transform | |
| CN108594834A (en) | One kind is towards more AUV adaptive targets search and barrier-avoiding method under circumstances not known | |
| CN108139225A (en) | Determine the layout information of motor vehicle | |
| CN109597417A (en) | A kind of more USV multi-agent synergy collision-avoidance planning methods based on collision prevention criterion | |
| CN108319293A (en) | A kind of UUV Realtime collision free planing methods based on LSTM networks | |
| CN109445440A (en) | The dynamic obstacle avoidance method with improvement Q learning algorithm is merged based on sensor | |
| Tan et al. | Gaussian process adaptive sampling using the cross-entropy method for environmental sensing and monitoring | |
| Lazarowska | Ant colony optimization based navigational decision support system | |
| CN111338356A (en) | Multi-target unmanned ship collision avoidance path planning method for improving distributed genetic algorithm | |
| He et al. | Dynamic adaptive intelligent navigation decision making method for multi-object situation in open water | |
| CN105000476A (en) | Unmanned crane space collision avoidance strategy based on fuzzy decision inference | |
| Wu et al. | Multi-vessels collision avoidance strategy for autonomous surface vehicles based on genetic algorithm in congested port environment | |
| CN114355923A (en) | MPC-based trajectory planning and tracking method under guidance of A |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190315 |
|
| WD01 | Invention patent application deemed withdrawn after publication |