U.S. patent application number 09/877493 was filed with the patent office on 2002-04-18 for track map generator.
Invention is credited to Cong, Shan.
Application Number | 20020044081 09/877493 |
Document ID | / |
Family ID | 26904923 |
Filed Date | 2002-04-18 |
United States Patent
Application |
20020044081 |
Kind Code |
A1 |
Cong, Shan |
April 18, 2002 |
Track map generator
Abstract
A radar system (14, 20, 22) generates a plurality of data points
(x, y) representative of the position of a tracked object (48), and
a representation of an associated path (50, 60) is formed
therefrom. At least one quality measure, and at least one heading
measure, of said representation, is calculated corresponding to at
least one map coordinate (44). The quality and heading measures are
stored in memory as a track map. Data from other tracked objects
(58) is used to update the track map (38), resulting a plurality of
heading values at associated map coordinates (44) that are
representative of a path (42) followed by the tracked objects (48,
58).
Inventors: |
Cong, Shan; (Ann Arbor,
MI) |
Correspondence
Address: |
DINNIN & DUNN P.C.
Attorneys and Counselors
Top of Troy Building
755 West Big Beaver Road
Troy
MI
48084
US
|
Family ID: |
26904923 |
Appl. No.: |
09/877493 |
Filed: |
June 8, 2001 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
60210193 |
Jun 8, 2000 |
|
|
|
60210878 |
Jun 9, 2000 |
|
|
|
Current U.S.
Class: |
342/70 ; 342/195;
342/90; 342/95; 701/301 |
Current CPC
Class: |
G01S 13/723 20130101;
G01S 13/931 20130101; G01S 2013/9329 20200101; G01S 2013/9322
20200101 |
Class at
Publication: |
342/70 ; 342/90;
342/95; 342/195; 701/301 |
International
Class: |
G01S 013/93 |
Claims
I claim:
1. A method of generating a track map, comprising: a. reading a
first plurality of first data points, wherein each first data point
of said first plurality of first data points is representative of a
position of first tracked object; b. generating a first
representation of a path of said first tracked object, from said
first plurality of first data points; c. calculating at least one
quality measure of said representation, wherein said at least one
quality measure corresponds to at least one map coordinate; d.
calculating at least one heading measure from said representation,
wherein said at least one heading measure corresponds to said at
least one map coordinate; and e. storing said at least one quality
measure and said at least one heading measures in a memory.
2. A method of generating a track map as recited in claim 1,
wherein said first representation of said path is a mathematical
representation selected from a linear fit of said first plurality
of first data points, a curve fit of said first plurality of first
data points, a Kalman smoothing of said first plurality of first
data points, and an autoregression of said first plurality of first
data points.
3. A method of generating a track map as recited in claim 1,
wherein said at least one quality measure is responsive to a
measure of fit by said representation of said first plurality of
first data points.
4. A method of generating a track map as recited in claim 1,
further comprising: a. reading a first plurality of second data
points, wherein each second data point of said first plurality of
second data points is representative of a velocity of first tracked
object, and said first plurality of second data points correspond
in time to said first plurality of first data points; and b.
calculating a first plurality of headings of said first tracked
object from said first plurality of second data points, wherein
said at least one quality measure is responsive to a heading error,
wherein said heading error is responsive to a difference between at
least one heading calculated from said first representation and at
least one heading of said first plurality of headings.
5. A method of generating a track map as recited in claim 1,
further comprising: a. reading a first plurality of third data
points, wherein each third data point of said first plurality of
third data points is representative of a heading of first tracked
object so that said first plurality of third data points are a
first plurality of headings, said first plurality of third data
points correspond in time to said first plurality of first data
points, wherein said heading error is responsive to a difference
between at least one heading calculated from said first
representation and at least one heading of said first plurality of
headings.
6. A method of generating a track map as recited in claim 1,
further comprising modifying at least one stored quality measure
responsive to at least one quality measure.
7. A method of generating a track map as recited in claim 1,
further comprising modifying at least one stored heading measure
responsive to at least one quality measure.
Description
[0001] The instant application claims the benefit of U.S.
Provisional Application Ser. No. 60/210,193 filed on Jun. 8, 2000
("ASL-264-PRO"), and of U.S. Provisional Application Ser. No.
60/210,878 filed on Jun. 9, 2000 ("ASL-266-PRO"), both of which
applications are incorporated herein by reference.
[0002] In the accompanying drawings:
[0003] FIG. 1 illustrates a block diagram of a radar processing
system that incorporates the instant invention;
[0004] FIG. 2 illustrates an example of a target tracking situation
during a first period of time;
[0005] FIG. 3 illustrates an example of a target tracking situation
during a second period of time after the first period of time of
FIG. 2;
[0006] FIG. 4 illustrates a general process in accordance with the
instant invention;
[0007] FIG. 5 illustrates a more detailed process in accordance
with the instant invention;
[0008] FIG. 6 illustrates a relationship of angular quantities;
[0009] FIG. 7 illustrates an example of a target tracking situation
during a first period of time, but with a host vehicle centered
track map; and
[0010] FIG. 8 illustrates an example of a target tracking situation
during a second period of time after the first period of time of
FIG. 7, but with a host vehicle centered track map.
[0011] There exists a need for an improved predictive collision
sensing or collision avoidance system for automotive applications
that can sense and identify an environment of a host vehicle with
sufficient range and accuracy so that proper countermeasures can be
selected and taken sufficiently early to either avoid a collision
or to mitigate injury therefrom either to occupants of the host
vehicle or to pedestrians thereto. As used herein, the term
predictive collision sensing system will also refer to a collision
avoidance system to mean a system that can sense and track targets
in the environment of the host vehicle, and then either suggest, or
automatically take countermeasures, that would improve safety.
Generally, a predictive collision sensing system tracks the motion
of the host vehicle relative to its environment, or vice versa, for
example, using a radar system with an associated target tracker.
The environment may include both stationary and moving targets. An
automotive environment is distinguished from other target tracking
environments--for example, that of air or sea vessels-in that
automotive vehicles are primarily operated in an environment that
is constrained by roadways. There are, of course, exceptions to
this, for example, parking lots or off-road driving conditions, but
these exceptions generally occupy a relatively small amount of
vehicular operating time, or a relatively small risk of collisions
that would benefit from a predictive collision sensing system.
[0012] If available, knowledge of roadway constraints, for example,
in the form of a map, can be useful in improving the performance of
a predictive crash sensing system. For example, if the host vehicle
is known to be operating on a particular roadway having a
particular path geometry, and an on-board navigation system detects
that the trajectory of the vehicle is departing from that path, for
example as a result of a skid or driver inattention, then the
predictive collision system could identify and react to this
situation, whether or not the on-board radar detected a potential
target with which the host vehicle might collide.
[0013] Knowledge of roadway constraints can also be used to improve
the convergence of an associated target tracker, or to improve an
estimate of the environment or situation in which the host vehicle
is operated.
[0014] Roadway constraints are generally characterized in the form
of a map of associated coordinates in a two-dimensional space. For
example, road trajectories can be plotted in an X-Y coordinate
system with positive X directed North, and positive Y directed
East. Whereas digitized road maps are presently widely available,
the extent of the utility thereof in a predictive crash sensing
system of a host vehicle is at least partially dependent upon a
navigation process to locate the host vehicle on the map assuming
that the maps are of sufficient. For example, the location and
direction of a vehicle can be measured by a GPS receiver; by a dead
reckoning system using measurements of vehicle heading from a
compass or directional gyroscope, and vehicle distance and heading
from wheel speed or rotation measurements, in conjunction with a
map matching algorithm; or a combination of the two. However,
heretofore available navigation systems are generally not
sufficiently accurate to provide a map of the roadway traveled by
the host vehicle, of sufficient accuracy for predictive crash
sensing. Moreover, road conditions can change over time, for
example, as a result of road construction, and these changes may
not always be timely entered into the associated digital maps.
[0015] Accordingly, there exists a need for an improved system and
method for generating a map of a track or tracks over which follow
the host vehicle and other vehicles in the environment of the host
vehicle.
[0016] Referring to FIGS. 1-3, a track map generator 10 is
illustrated in a radar processing system 12 for processing radar
data from a radar system 14 incorporated in a host vehicle 16. is
equipped with a sensor for sensing targets in the environment
thereof and for actuating and/or controlling associated
countermeasures responsive to the relative motion of the host
vehicle and one or more targets. The sensor for sensing targets is
for example a radar or lidar sensor system that senses and tracks
the location of targets relative to the host vehicle, and predicts
if a collision between the host vehicle and the target is likely to
occur, for example, as disclosed in U.S. Pat. No. 6,085,151,
assigned to the assignee of the instant invention, and incorporated
by reference herein.
[0017] For example, referring to FIG. 1, illustrating a block
diagram of a radar processing system 12, each block in the block
diagram comprises associated software modules that receive, prepare
and/or process data provided by a radar system 14 mounted in the
host vehicle. Data from the radar system 14 is preprocessed by a
preprocessor 18 so as to generate radar data, for example, range,
range rate, azimuth angle, and quality thereof, suitable for target
tracking. The radar data may typically covers a wide field of view
forward of the host vehicle, at least .+-.5 degrees from the host
vehicle's longitudinal axis, and possibly extending to .+-.180
degrees or larger depending upon the radar and associated antenna
configuration. A present exemplary system has a field of view of
+/-55 degrees.
[0018] A tracker 20 converts radar output data (range, range rate
& azimuth angle) into target speed and x-y coordinates
specifying the location of a target. For example, the system of
Application '035 discloses a system for tracking multiple targets,
and for clustering associated radar data for a single target. The
associator 22 relates older track data to that from the latest
scan, compiling a track history of each target.
[0019] The track map generator 10--described more fully
hereinbelow--stores this data, and generates a track map therefrom
as a record of the progress of target motions relative to the host
vehicle.
[0020] A situation awareness processor 24 uses 1) the track map, 2)
data acquired indicating the motion of the host vehicle, i.e. host
vehicle information 26, and possibly 3) "environmental data", to
determine the most likely or appropriate driving situation from a
set of possible driving situations. For example, the "environmental
data" can include GPS data, digital maps, real-time radio inputs of
highway geometry and nearby vehicles, and data from real-time
transponders such as electromagnetic or optical markers built into
highways.
[0021] For example, the "environmental data" can be used by a road
curvature estimator 28 to provide an estimate of road curvature to
the situation awareness processor 24.
[0022] The situation awareness processor 24 stores and interprets
the track map from the track map generator 10, and compares the
progress over time of several target tracks. Evaluation of the
relative positions and progress of the tracked targets permits
identification of various driving situations, for example a
location situation, a traffic situation, a driving maneuver
situation, or the occurrence of sudden events.
[0023] The situation estimated by the situation awareness processor
24, together with collision and crash severity estimates from a
collision estimator 30 and a crash severity estimator 32
respectively, are used as inputs to a response generator 34 to
select an appropriate countermeasure 36 for example, using a
decision matrix. The decision of a particular response by the
response generator 34 may be based on, for example, a rule-based
system (an expert system), a neural network, or another decision
means.
[0024] Examples of countermeasures 36 that can be activated
include, a warning device to warn the driver to take corrective
action, for example 3D audio warning (for example, as disclosed in
U.S. Pat. No. 5,979,586 assigned to the assignee of the instant
invention and incorporated by reference herein); various means for
taking evasive action to avoid a collision, for example the engine
throttle, the vehicle transmission, the vehicle braking system, or
the vehicle steering system; and various means for mitigating
injury to an occupant if a collision is unavoidable, for example a
motorized safety belt pretensioner, or internal or external
airbags. The particular one or more countermeasures 36 selected,
and the manner by which that one or more countermeasures 36 are
activated, actuated, or controlled, depends up the situation
identified by the situation awareness processor 24, and upon the
collision and crash severity estimates. By way of example, one
potential scenario is that the response to encroachment into the
host's lane of travel requires a different response depending upon
whether the target is coming from the opposite direction or going
the same way as the host vehicle, but cutting into the host's lane.
By considering the traffic situation giving rise to the threat, the
countermeasures 36 can be better adapted to mitigating that threat.
By using a radar system, or generally a predictive collision
sensing system, to sense targets within range of the host vehicle,
the countermeasures 36 may be implemented prior to an actual
collision so as to either avoid the collision or to mitigate
occupant injury from the collision.
[0025] The track map generator 10 operates on an assumption that
objects in the field of view of a tracking sensor are subjected to
evolving but common path constraints, so that, although they have
independent controls, their associated trajectories are correlated
with one another. This assumption is generally true for ground
target trajectory estimation, where most likely the targets are
motor vehicles and are driving on roads where there are strong path
restrictions. In such cases, using the constraints will certainly
improve both target kinematic states tracking performance and
target status estimation performance. Here target status refers to
decisions such as the relative position of a target on a road and
if a target is making an abnormal maneuver.
[0026] Generally a road map is not available that can be updated in
real-time to provide instant road condition report, nor can
sufficient computation resources be allocated so as to provide a
road map in real-time. This limitation is overcome by providing a
dynamic map of previous tracks can be used as an a map. A map
constructed in this way would not provide information of the first
object passing by the sensor, however, as more and more targets are
encountered, the accuracy of the map improves, so as to be able to
provide more significant information. This map is referred to as a
track map 38 because it is built by accumulating previous target
tracks.
[0027] A track map 38 can be used in a number of ways. For example,
by accumulating information in previously existing tracks 40, the
convergence speed of a new track can be improved and tracking error
at the early stage of a track can be reduced. As another example,
as track information is combined into a map, vital information
about the current and future traffic situation can also be
deduced.
[0028] A particular track 40 follows a path 42, and can be
characterized by the content and quality of the associated
information. The information content is the proposition that at
location L a target has states S, and the information quality is
the strength with which that proposition is believed to be true.
The he information content can have different forms, for example,
including but not limited to target speed, heading, acceleration
and type. The associated information quality can be evaluated under
various reasoning frameworks, such as probability, fuzzy logic,
evidential reasoning, or random set.
[0029] For example, the track map 38 comprises a Cartesian grid of
individual cells located by coordinates i and j, each cell
44--denoted grid(i,j)--has associated state and quality information
as follows: grid(ij).states, and grid(ij).quality. The resolution
and size of the track map 38 is dependent upon the associated
computing bandwidth of an associated processor used to generate the
track map.
[0030] A path 42 is then represented as a collection of grids:
{grid(ij)}. Each time a track is obtained, its associated path 42
can also be deduced. A newly recognized path 42 can then be
registered into the track map 38 and the associated information
content and quality of the map are also updated. This procedure is
called map building. Once a map has been built up, it can be used
to improve tracks 40 obtained thereafter. This procedure is called
track-map fusion.
[0031] Referring to FIG. 4, illustrating a map building process, in
step (402) the track data is read from the tracker 20/associator
22. In step (404), the backprojected path of the track is
determined, which is then related to the map grid of the track map
38 in step (406). Then in step (408) the map grid is updated, after
which the process is repeated in step (402).
[0032] Referring to FIG. 5, illustrating the map building and
fusion processes in greater detail, in step (502) the track data
comprising position (x,y), velocity (Vx, Vy), target angle ang_h,
and data quality S.sub.t is read from the tracker 20/associator 22.
If, in step (504), the track is new, then in step (506), the track
fit data is initialize. For example, accumulators of summation
processes associated with the fitting processes are initialized to
zero before accumulating new track data thereinto. Otherwise, in
step (508), the new track data is accumulated in the associated
accumulators used in the fitting process. Then in step (508)--which
corresponds to step (404) of FIG. 4, the track history data is
processed --for example by a smoothing, regression, or other curve
fitting process--to generate a representation of the path of the
track.
[0033] Referring in greater detail to steps (404) of FIG. 4, and
(508) of FIG. 5, generally the backprojected path 42 of a track 40
comprises smoothed trajectory of an existing track 40. As the
tracks are obtained from a stochastic calculation, at a specific
time an existing track 40 represents the estimated target states
based on the information up to that moment. When more information
about the target arrives, a better estimate can be derived about
the earlier moment. The backprojected path 42 is the improved the
history of target states. To obtain a backprojected path 42, one
may use Kalman smoother, autoregression, or a line or curve fitting
process. The particular approach depends on a tradeoff between
nature of target model, performance, available computation
bandwidth.
[0034] For example, autoregression would be appropriate if the
backprojected path 42 is subject to strong trajectory restriction
and has a smooth trajectory. A typical curve function in parametric
form is:
y=f(x)=a.sub.nx.sup.n+a.sub.n-1x.sup.n-1+. . .
+a.sub.1x+a.sub.0
[0035] where {a.sub.1,i=0,1, . . . n} are the results of
regression, and jointly define a curve.
[0036] As another example, the backprojected path may be found by
one-dimensional line fitting of the associated track history, as
follows: 1 S x ( k ) = j = 1 k x j f fit j - k = S x ( k - 1 ) f
fit + x k S y ( k ) = j = 1 k y j f fit j - k = S y ( k - 1 ) f fit
+ y k S xy ( k ) = j = 1 k x j y j f fit j - k = S xy ( k - 1 ) f
fit + x k y k S x 2 ( k ) = j = 1 k x j 2 f fit j - k = S x2 ( k -
1 ) f fit + x k 2 S w ( k ) = j = 1 k f fit j - k = S w ( k - 1 ) f
fit + 1 a = ( S x S y - S w S xy ) / ( S x 2 - S w S x 2 ) b = ( S
x S xy - S x2 S y ) / ( S x 2 - S w S x 2 ) d x = x 1 - x k d y = y
1 - y k
[0037] where a and b are the fitting results which jointly define a
straight line pattern:
y=ax+b
[0038] and dx and dy define the distance from the starting point to
the current point. Also note that a fading factor is added to
gradually decrease the contribution of the early part of a track to
current fitting result, for example:
f.sub.fit=0.98
[0039] Given a backprojected target trajectory y=f(x),x.di-elect
cons.[x.sub.1,x.sub.N] from either steps (406) or step (510), where
x.sub.1.ltoreq.x.sub.N, in steps (406) and (512) the path 42 of the
backprojected target trajectory is associated with the particular
cells 44 of the track map 38. For the size of each cell 44 given as
.DELTA..sub.x and .DELTA..sub.y for x and y dimensions
respectively, and for the center of grid(0,0) starting at
(x.sub.g0, y.sub.g0), then the starting and ending cell locations
are given by:
x.sub.g1=mod(x.sub.1-x.sub.g0,.DELTA..sub.x).DELTA..sub.x+x.sub.g0
and
y.sub.g1=mod(y.sub.1-y.sub.g0,.DELTA..sub.y).DELTA..sub.y+y.sub.g0
x.sub.gN=mod(x.sub.N-x.sub.g0,.DELTA..sub.x).DELTA..sub.x+x.sub.g0
and
y.sub.gN=mod(y.sub.N-y.sub.g0,.DELTA..sub.y).DELTA..sub.y+y.sub.g0
[0040] If .vertline.df(x)/dx.vertline..ltoreq.1, cell locations
between the starting and ending cells 44 are found by
x.sub.i=x.sub.1+i.DELTA..sub.x,y.sub.i=f(x.sub.i)
[0041] and
x.sub.gi=mod(x.sub.i-x.sub.g0,.DELTA..sub.x).DELTA..sub.x+x.sub.g0
and
y.sub.gi=mod(y.sub.i-y.sub.g0,.DELTA..sub.y).DELTA..sub.y+y.sub.g0
[0042] otherwise,
y.sub.i=y.sub.1+i.DELTA..sub.y,x.sub.i=f.sup.-1(y.sub.i)
[0043] and
x.sub.gi=mod(x.sub.i-x.sub.g0,.DELTA..sub.x).DELTA..sub.x+x.sub.g0
and
y.sub.gi=mod(y.sub.i-y.sub.g0,.DELTA..sub.y).DELTA..sub.y+y.sub.g0
[0044] The track map 38 is updated by first calculating the
associated quality measures in steps (514), (516) and (518) as
follows:
[0045] Fitting quality measure of step (518) contains two parts,
one is for the contradiction of curve fitting result from step
(510) and the output of the tracker 20/associator 22 from step
(502); and one is for the difference between a fitted curve (or
path 42) and a track 40, i.e.,
M.sub.fit=M.sub.contM.sub.diff
[0046] M.sub.diff is defined by the distance between the newly
found curve and the existing track:
M.sub.diff=f({.parallel.{circumflex over
(X)}.sub.k-X.sub.k.parallel.,k=1,- 2, . . . N})
[0047] where q.sub.trk is the quality of current track,
.parallel..cndot..parallel. is the distance between a point on a
track at time k, X.sub.k, and the corresponding backprojected point
{circumflex over (X)}.sub.k, N is the number of points in a track.
The simplest form of the distance is an Euclid distance: 2 ; X ^ k
- X k r; = ; ( i = 0 n a i x k i - y k ) 2 r;
[0048] Depending upon the particular situation, the function
f({.parallel.{circumflex over (X)}.sub.k-X.sub.k.parallel.,k=1,2, .
. . N}) can be defined as any monotonic function with output
between 0 and 1, as long as it consistently conveys the quality of
the fitting and has a set of well defined operations.
[0049] M.sub.cont denotes contradiction component of quality
corresponding to the discrepancy between the fitting result and and
the output of the tracker 20/associator 22 from step (502). In each
cell 44, the contradiction is reflected by the difference between
target heading angle given by the tracker 20/associator 22 and the
target heading angle defined by the tangent of the curve at the
center of the grid, i.e., 3 f ( ang_f , ang_h ) = { ang_f - ang_h ,
if ang_f - ang_h < 2 - ang_f - ang_h , if ang_f - ang_h
[0050] where
M.sub.cont=f(ang.sub.--f,ang.sub.--h)
[0051] As in the case of M.sub.diff, f(ang_f,ang_h) can be defined
as any monotonic function between 0 and 1, as long as the
difference between the two angle is reflected.
[0052] Finally, before the newly obtained information can be used
for updating each grid, one must check the quality of the curve
fitting. The this end, each fitting has to consider the following:
a) the track should be long enough to guarantee a meaningful
fitting; b) the fitting should be well-defined for numerical
stability; and c) M.sub.diff itself should be small enough to
guarantee a good fitting quality.
[0053] As a particular example, the contradiction component of
quality can be given by:
M.sub.cont=s,e.sup.-2f(ang.sup..sub.--.sup.f,ang.sup..sub.--.sup.h)
[0054] Where 4 ang_h = a tan 2 ( v y , v x ) ang_f = { a tan ( f (
x ) x x = x g ) , if a tan ( f ( x ) x x = x g ) - ang_h < 2 a
tan ( f ( x ) x x = x g ) + , if a tan ( f ( x ) x x = x g ) -
ang_h 2
[0055] and S.sub.t is a quality measure from the tracker
20/associator 22 of the current track, ang_f and ang_h are the
estimates of the target heading angle obtained from line fitting
and the tracker 20/associator 22, respectively. ang_f and
ang.sub.--h are defined as: 5 ang_h = a tan 2 ( v y , v x ) ang_f =
{ a tan ( a ) , if a tan ( a ) - ang_h < 2 a tan ( a ) + , if a
tan ( a ) - ang_h 2
[0056] wherein vx and vy are estimated target velocities on x and y
directions, and the angular relationships are illustrated in FIG.
6.
[0057] The difference between a fitted line and the original track:
6 diff = i = k - 2 k ( x fi - x i ) 2 + ( y fi - y i ) 2 M diff = e
- diff / 4
[0058] where
(x.sub.fi,y.sub.fi) and (x.sub.1, y.sub.l)
[0059] are points on a fitted line and a track, respectively. Here
to save computation bandwidth, we use the most recent three points
to measure the difference.
[0060] The resulting fitting quality measure is the multiplication
of Mcont and Mdiff:
M.sub.fit=M.sub.contM.sub.diff
[0061] Prior to updating the cells 44 with new information, the old
information therein is aged, or faded, in step (520), so as to
reduce the significance of cells 44 for which tracking data is no
longer measured, and to eventually clear out old data from the
cells 44. For example, this may be accomplished by multiplying the
information in each grid by a fading factor f.sub.d as follows:
grid(i,j).quality=f.sub.d(i,j).times.grid(i,j).quality
[0062] wherein the fading factor is between zero and one so that
the quality of each grid can be degraded to zero if no new
information is received therein.
[0063] For example, a fading factor may be given by:
f.sub.d=0.14e.sup.-(n.sup..sub.r.sup.+1)/(.vertline.v.vertline.+6)+0.85
[0064] where n.sub.r is the number of reports and v is the
velocity, if available. The numbers, such as 0.14, 0.85 and 6, are
given to tune the decreasing speed.
[0065] The track map 38 is then updated in steps (522) and (524) by
a backprojection process that updates the grids from starting point
to current point. This updating process may be subject to
conditions, for example,
[0066] a) that the track has existed for more than six scans
(wherein track length --the number of times a track has been
updated--is used as an additional measure for association quality
and filtering quality);
[0067] b) that the fitting algorithm is well defined, i.e.,
.vertline.S.sub.x.sup.2-S.sub.wS.sub.x2.vertline.>0
[0068] c) that diff<10 so that the fit is of relatively good
quality; or
[0069] d) that the track is long enough to establish a meaningful
fitting, i.e.,
max(.vertline.d.sub.x.vertline.,.vertline.d.sub.y.vertline.)<2
[0070] To update a grid, first the existing information in the grid
is compared with the newly obtained information, as follows:
p.sub.cont=1-p.sub.ge.sup.31
f.sup..sub.c.sup.(ang.sup..sub.13.sup.g,ang.s- up..sub.13
.sup.f)
[0071] where p.sub.g is the quality of the current information
content in this grid, and 7 f c ( ang_g , ang_f ) = { ang_f - ang_g
, if ang_f - ang_g 2 - ang_f - ang_g , if ang_f - ang_g >
[0072] The information quality is normalized by using:
p.sub.cur=M.sub.fit/(M.sub.fit+p.sub.cont+0.05)
[0073] where 0.05 accounts for unknowns and for numerical
stability, and p.sub.cur defines the quality for newly obtained
information. With p.sub.g representing the quality of information
content in the original grid, then the information in the cells 44
is updated as follows (Bayesian):
[0074] (a) Heading:
grid.heading=(grid.heading p.sub.cur+ang.sub.--f
p.sub.g)/(p.sub.cur+p.sub- .g)
[0075] (b) Quality:
grid.quality=p.sub.gp.sub.cur/(p.sub.gp.sub.cur+(1-p.sub.cur)(1-P.sub.g))
[0076] Forward projection is used to look ahead of the current
track to the region where this target has no history and to revive
the history of the previous tracks. Basically the same technique
used in 3.3 can be applied here. The difference lies in the stop
condition. In back projection, well defined starting and ending
points exist. However, in forward projection, no clearly defined
ending point of the projection exists. Instead, we introduce the
following two conditions:
[0077] (a) Stop Condition 1: Reach a point with no information,
i.e.
grid.quality.ltoreq..xi., .xi..gtoreq.0.
[0078] (b) Stop Condition 2: Reach a grid where the information
content is too different with the projection, i.e.
p.sub.cont.gtoreq.T,T.gtoreq.0.
[0079] Here T serves as a threshold
[0080] The operation of the track map generator 10 is illustrated
in FIGS. 2 and 3. Referring to FIG. 2, a host vehicle 16 at first
position 16.1 travels over a first trajectory 46 over a first
period of time to a second position 16.2. Simultaneously, a first
target vehicle 48 at first position 48.1 travels over a second
trajectory 50 over the first period of time to a second position
48.2. The track of the first target vehicle 48 is measured by the
radar system 14 and associated elements of the radar processing
system 12 at associated sampling times 52. The second trajectory 50
intersects a first set of cells 54 that are updated by the track
map generator 10.
[0081] Referring to FIG. 3, during a second period of time, the
host vehicle 16 travels from the second position 16.2 to a third
position 16.3 over the first trajectory 46, and simultaneously, a
second target vehicle 58 at first position 58.1 travels over a
third trajectory 60 to a second position 58.2. The track of the
second target vehicle 58 is measured by the radar system 14 and
associated elements of the radar processing system 12 at associated
sampling times 54. The third trajectory 60 intersects a second set
of cells 62, some of which are also of the first set of cells 56,
that are updated by the track map generator 10. The resulting
updated cells each contain an associated direction and quality
representing a composite of the paths of the first 48 and second 58
target vehicles, relative to the host vehicle 16.
[0082] FIGS. 2 and 3 illustrate track maps 38 with associated cells
that are fixed in space, as may be suitable for used in conjunction
with absolute navigation data. Referring to FIGS. 7 and 8--which
correspond to the situations illustrated in FIGS. 2 and 3--the
coordinate system of the track map may also be adapted to move with
the host vehicle 16, so as to provide a map that is localized to
the host vehicle 16.
[0083] While specific embodiments have been described in detail in
the foregoing detailed description and illustrated in the
accompanying drawings, those with ordinary skill in the art will
appreciate that various modifications and alternatives to those
details could be developed in light of the overall teachings of the
disclosure. Accordingly, the particular arrangements disclosed are
meant to be illustrative only and not limiting as to the scope of
the invention, which is to be given the full breadth of the
appended claims and any and all equivalents thereof.
* * * * *