Sensor fusion system and vehicle control system therewith
Each of multiple probability distribution outputting units computes a probability distribution of a data value detected by the corresponding sensor or algorithm in an image recognition processing or the like. The respective probability distributions of the multiple probability distribution outputting units are given as outputs to a synthetic determination processing unit. Data formats of the outputs to the synthetic determination processing unit can be thereby standardized. Hence, the synthetic determination processing unit is exempted from considering which type of sensor or algorithm each of the outputs is based upon. Even when a sensor or algorithm is added or changed, the same data-fusing algorithm in the synthetic determination processing unit can be uniformly used.
1. A sensor fusion system comprising:
a plurality of detecting units, each of which detects a data value of a given data kind;
a plurality of probability distribution computing units, each of which computes a probability distribution of the data value detected by a corresponding detecting unit included in the plurality of detecting units, wherein the probability distribution of the data value indicates likelihood to a true value of the given data kind, and then outputs the computed probability distribution; and
a synthetic determining unit that estimates a most likely data value of the given data kind based on the probability distributions outputted from the plurality of probability distribution computing units, wherein
each of the plurality of probability distribution computing units comprises a probability distribution converting unit, which converts the data value detected by the corresponding detecting unit into the probability distribution by using distribution data statistically aggregated from data values detected in past times by the corresponding detecting unit; and
the probability distribution computed by each of the probability distribution computing units is represented as a sum of a Gaussian distribution and a uniform distribution.
2. The sensor fusion system of claim 1 ,
wherein at least one of the probability distribution computing units converts a shape of the computed probability distribution to at least one scalar value and outputs the at least one scalar value,
the at least one scalar value including one of (i) an average of the Gaussian distribution, (ii) a variance of the Gaussian distribution, and (iii) a normal recognition ratio indicating a ratio between the Gaussian distribution and the uniform distribution, and
wherein the synthetic determining unit converts the at least one scalar value to the shape of the probability distribution.
3. The sensor fusion system of claim 1 ,
wherein at least one of the probability distribution computing units receives information relating to a parameter affecting detection accuracy in the corresponding detecting unit and changes the probability distribution based on the received information.
4. The sensor fusion system of claim 1 , wherein each of the probability distribution computing units computes as an output, the probability distribution which is a likelihood statistically computed from previous data values, provided to each of the probability distribution computing units.
5. The sensor fusion system of claim 1 ,
wherein the uniform distribution is based on a mis-recognition probability obtained from (i) an actual detection result or (ii) a definition or property of each of the detecting units.
6. The sensor fusion system of claim 1 , wherein the statistically aggregated distribution data shows a relationship between a true value of a corresponding data kind and the data values detected in the past times by the corresponding detecting unit.
7. A vehicle control system mounted in a vehicle, comprising:
a sensor fusion system that includes
a plurality of detecting units, each of which detects a data value of a given data kind,
a plurality of probability distribution computing units, each of which computes a probability distribution of the data value detected by a corresponding detecting unit included in the plurality of detecting units, wherein the probability distribution of the data value indicates a likelihood to a true value of the given data kind, and then outputs the computed probability distribution, and
a synthetic determining unit that estimates a most likely data value of the given data kind based on the probability distributions outputted from the plurality of probability distribution computing units, wherein
each of the plurality of probability distribution computing units comprises a probability distribution converting unit, which converts the data value detected by the corresponding detecting unit into the probability distribution by using distribution data statistically aggregated from data values detected in past times by the corresponding detecting unit;
at least one control unit that receives the estimated most likely data value that is given from the synthetic determining unit and controls an in-vehicle device based on the received most likely data value of the given data kind; and
the probability distribution computed by the each of the probability distribution computing units is represented as a sum of a Gaussian distribution and a uniform distribution.
8. The vehicle control system of claim 7 ,
wherein the synthetic determining unit estimates the most likely data value of the given data kind that includes at least one of a road surface state, a solar radiation amount, a position of an object, a speed of the object, a size of the object, a type of the object, wherein the object includes a preceding vehicle, and
wherein the at least one control unit includes at least one of
a control unit of an information providing device for providing information to a driver of the vehicle;
a control unit of a warning device,
a control unit of a traveling device for controlling a vehicle traveling state,
a control unit of a driving-state control device for controlling a state of driving operation device, and
a control unit of an occupant protecting device for protecting an occupant.
9. The vehicle control system of claim 7 ,
wherein, when the synthetic determining unit gives the estimated most likely data value to the at least one control unit, the synthetic determining unit also gives reliability data indicating reliability in the estimated most likely data value.
10. The vehicle control system of claim 9 ,
wherein the reliability data includes at least one of
an error probability in the estimated most likely data value,
a correct answer probability in the estimated most likely data value,
a probability distribution in the estimated most likely data value,
an average of the probability distribution,
a variance of the probability distribution, and
a normal recognition ratio.
11. The vehicle control system of claim 7 ,
wherein at least one of the probability distribution computing units converts a shape of the computed probability distribution to at least one scalar value and outputs the at least one scalar value,
the at least one scalar value including one of (i) an average of the Gaussian distribution, (ii) a variance of the Gaussian distribution, and (iii) a normal recognition ratio indicating a ratio between the Gaussian distribution and the uniform distribution, and
wherein the synthetic determining unit converts the at least one scalar value to the shape of the probability distribution.
12. The vehicle control system of claim 7 ,
wherein at least one of the probability distribution computing units receives information relating to a parameter affecting detection accuracy in the corresponding detecting unit and changes the probability distribution based on the received information.
13. The sensor fusion system of claim 7 , wherein the statistically aggregated distribution data shows a relationship between a true value of a corresponding data kind and the data values detected in the past times by the corresponding detecting unit.