Patent Application
20140358413
Not Applicable.
Not Applicable.
The present invention relates in general to monitoring traffic surrounding a motor vehicle, and, more specifically, to a method and apparatus for classifying on-board and in real time a traffic density within which a host vehicle is moving.
For a variety of automotive systems and functions, it can be useful to have available an estimate of local traffic density (including estimations of traffic density in the direct forward path of the vehicle, in adjacent lanes, and an aggregate or overall traffic density in the vicinity of the vehicle). For example, the warning thresholds (e.g., distances or buffer zones) for a collision warning system may be adjusted depending on whether traffic density is light, medium, or heavy. In addition, a driver alertness monitoring system may use different thresholds according to the traffic density.
Conventionally, traffic density estimations have been obtained in various ways. In one automated technique, a rough estimate of traffic density is found by tracking cell phones passing through designated roadway locations (e.g., a central monitor obtains GPS or cell tower-based coordinates of individual phones, maps them onto roadway segments, calculates a vehicle density, and communicates the result to the vehicles). Other automated techniques for counting the number of vehicles present at a road segment can also be used. These approaches give only a general idea of how many vehicles are within a fixed area (i.e., not specific to the immediate area around any particular vehicle). They have other disadvantages including that the update rate is slow, the vehicle must have wireless communication in order to access the information, and infrastructure must be provided for performing the calculations outside of the host vehicle.
In another approach, drivers or other observers may visually characterize the amount of traffic in an area. This is subject to the same disadvantages, and may be less accurate. In yet another approach, a Vehicle-to-Infrastructure system may be used to characterize the traffic density. This is subject to high costs of implementing hardware on both the vehicles and the roadside. Additionally, a sufficient market penetration would be needed in order for this to be feasible.
In one aspect of the invention, a method is provided for an electronic controller in a host vehicle to determine a traffic density. A sensor remotely senses objects within a field of view around the host vehicle. Positions are identified of nearby vehicles within the sensed objects. A path of a host lane being driven by the host vehicle is predicted. The electronic controller bins the nearby vehicles into a plurality of lanes including the host lane and one or more adjacent lanes flanking the predicted path. The electronic controller determines a host lane distance in response to a position of a farthest vehicle that is binned to the host lane, and then determines an adjacent lane distance in response to a difference between a closest position in an adjacent lane that is within the field of view and a position of a farthest vehicle binned to the adjacent lane. The electronic controller indicates a traffic density in response to a ratio between a count of the binned vehicles and a sum of the distances.
In a preferred embodiment, the vehicle locations on the surrounding roadway are estimated through the use of an on-board forward looking sensor. Additional vehicle sensors such as side looking blind spot sensors or rear looking sensors can also be used.
The relative positions of nearby vehicles (laterally and longitudinally) are acquired from the forward looking sensor. This can be either directly in Cartesian form or calculated from polar coordinates. All of the target vehicles that are detected by the forward looking sensor are then be binned into “lanes” based on their offset from the predicted path of the host vehicle. The predicted path may be determined from a yaw rate sensor or GPS Map data, for example. Based on a typical lane width, the host lane is considered to occupy an area +/− one-half of the lane width around the predicted path. An adjacent lane to the right of the host measured from the host's center line goes from +½ lane width to +1½ lane width, while an adjacent lane to the left measured from the host's center line goes from −½ lane width to −1½ lane width. This calculation can be carried out to any desired number of total lanes of interest.
With the vehicles all binned to lanes, a count is then performed to determine the total number of vehicles that are seen in each lane. For the host vehicle's lane, the count should include the host vehicle. To complete a density calculation, a value for the monitored distance within each lane is needed. For the host's lane, this is done by determining which vehicle is the farthest forward in the host's lane. The length of the host vehicle and an estimate of the most forward vehicle's length are preferably added to the longitudinal relative position measured from the front of the host vehicle to the rear of the most forward in lane vehicle to yield a longitudinal distance in which vehicles are seen for the host's lane. If no forward vehicles are seen, then the distance may default to the maximum reliable detection distance of the sensor.
For the adjacent lanes, a distance is preferably determined in response to the field of view from the location of the forward looking sensor to determine the closest point to the host vehicle that a vehicle in the adjacent lane could be detected. This detection distance is then subtracted from the longitudinal relative position of the most forward vehicle in the adjacent lane (preferably again adding a length estimate for the detected vehicle and defaulting to a maximum detection distance if no vehicles are found). The ratio of each respective count to the respective detection distance gives the traffic density for the respective lane. An overall density is obtained from the ratio of the total count to the summed distances.
Referring now to
In operation, traffic density controller 26 identifies a predicted path of the host vehicle in one of several ways. For example, an optically-based lane detection system 27 coupled to camera 24 may employ pattern recognition to detect lane markers or other features to locate the roadway lanes. Thus, the paths of the host lane and adjacent lanes may be fed directly to controller 26. Alternatively, a vehicle yaw sensor 28 may be coupled to controller 26 for providing lateral acceleration information to be used by controller 26 to predict the lane path. In another alternative, a GPS navigation/mapping system 30 may be coupled to controller 26 for identifying lane locations based on using detected geographic coordinates of host vehicle 11 as a pointer onto a roadway map.
Based upon vehicle counts and lane distances as determined below, controller 26 generates traffic density indications for the purpose of providing them to other appropriate controllers (not shown) and/or functions that modify their performance in accordance with the traffic density. The indications may be communicated within the vehicle over a multiplex bus 31. Based on the indicated traffic density, the other systems may adjust thresholds or other aspects of their system operation to account for the actual traffic conditions determined in the immediate vicinity of the host vehicle on-board and in real-time.
As shown in
Once the host and adjacent lanes have been laid out relative to the position of the host vehicle, each tracked vehicle can be binned according to the areas covered by the lanes.
With the count information obtained, the next step is to derive the roadway distances over which the counted vehicles are distributed. Within the field of view of the remote sensors, there is a maximum detection distance for sensing any vehicles that may be present. Whenever vehicles are present, however, the view out to the maximum distance may be blocked by a detected vehicle. In the example of
In an adjacent lane to the side of host vehicle 35, the appropriate distance to be used as a basis for the density calculation usually does not begin at a point even with the host vehicle because the field of view for the sensing system is unlikely to correspond with the exact front of host vehicle 35. When using just a forward-looking detector, a vehicle in an adjacent lane must be at least slightly ahead of host vehicle 35 in order to be detected. Locations 46 and 47 in the adjacent lanes correspond to a closest position in those adjacent lanes that is within the field of view of the sensors. These locations can be measured in advance during the design of the vehicle.
For an object detection system with other types of sensors, the beginning position for the distance measurement can be at other positions relative to the host vehicle. For detectors with side-looking sensors or rear-looking sensors, the starting position for determining adjacent lane distances could even be behind host vehicle 35 or could be defined according to a farthest detected adjacent vehicle behind the host vehicle.
For right adjacent lane 40, the adjacent lane distance to be used in the traffic density calculation comprises a range R5 between position 47 and a farthest vehicle 42 in lane 40 plus a length L3 corresponding to the type of vehicle identified by the object tracking system (e.g., a representative car or truck length). Similarly, a distance for adjacent lane 37 comprises a range R3 between position 46 and vehicle 38 plus an incremental length L2 of vehicle 38 (either estimated or measured).
In step 54, the furthest ahead vehicle is found for each lane having a vehicle present. For the host lane, this distance along with the host length and furthest vehicle length is used to derive the distance over which vehicles in the lane are distributed. For the adjacent lanes, it is the furthest vehicle and length in combination with the closest detectable point in the lane that is used. If no vehicles are present in a lane, then the associated distance defaults to a maximum detection distance of the sensors along the predicted path of the respective lane. This predetermined maximum detection distance may be a fixed value stored in the controller or could be calculated based on environmental factors such as the height of the horizon. In step 55, a density is calculated for each lane equal to the respective vehicle count divided by the distance determined for each respective lane. In step 56, an overall density equal to the total count divided by the sum of distances is determined.
The raw traffic density values obtained in steps 55 and 56 can be directly used, or the raw values may be normalized or classified in step 57. Normalizing may preferably be comprised of transforming the values onto a scale between 0 and 1, determined as a percentage of a predetermined heavy traffic density threshold. For example, a raw value for an overall traffic density would be divided by the threshold and then clipped to a maximum value of 1. The predetermined heavy threshold may be empirically derived based on the prevalent traffic conditions in the market where the vehicle is to be sold and used.
Alternatively, classifying the raw traffic density values may be comprised of defining light, medium, and heavy traffic thresholds. Depending on the range in which the raw traffic density values fall, the corresponding level of light, medium, or heavy traffic density could be determined and reported to the other vehicle systems. Thus, the traffic density value or values, whether raw, normalized, or classified, are indicated to the appropriate functions or systems that need them in step 58.
Preferably the method of the invention may be performed using only valid lanes that can be verified to exist around the host vehicle as shown in
To identify valid lanes, the method in
If no vehicles are detected in the currently-examined lane in step 61, then the method proceeds in step 63 wherein the present overall traffic density is used to determine a time value Y. In situations where a higher traffic density exists, the likelihood of an empty lane is reduced. In conditions of a light traffic density, the possibility of a valid lane being empty of vehicles for a longer period of time increases. Therefore, a time value Y is selected with a magnitude that reflects an average wait time during which it would be expected that a vehicle would again appear in the empty lane. In step 64, a check is made to determine whether the potential lane being checked has been empty for the last Y seconds. If not, then the lane is still considered valid and a return is made to step 60. If the lane has been empty for Y seconds, then it is not considered a valid lane in step 65. The invalid lane may typically be excluded from the density calculations until a vehicle is detected in that potential lane.
