System and method for aligning sensors on a vehicle

Description

RELATED FILINGS

This application is a continuation application of Ser. No. 10/985,577 filed Nov. 9, 2004 now U.S. Pat. No. 7,337,650 which is incorporated herein by reference.

This application incorporates by reference U.S. Pat. Nos. 6,629,033, Issued Sep. 30, 2003 Titled—OPEN COMMUNICATION SYSTEM FOR REAL-TIME MULTIPROCESSOR APPLICATIONS, 6,771,208, Issued Aug. 3, 2004 Titled—MULTI SENSOR SYSTEM, and U.S. patent application Ser. No. 09/841,915, filed Apr. 24, 2001 entitled: METHOD AND APPARATUS FOR DYNAMIC CONFIGURATION OF MULTIPROCESSOR SYSTEM.

REFERENCES CITED

U.S. Patent Documents

• U.S. Pat. No. 5,245,909 May 7, 1990; Title: Automatic Sensor Alignment; Inventors: Corrigan; John D. (Bridgeton, Mo.); Simmons, Jr.; Dennis E. (St. Louis, Mo.); Berg; Robert L. (St. Louis, Mo.); Assignee: McDonnell Douglas Corporation (MD)
• U.S. Pat. No. 6,202,027 Mar. 13, 2001; Title: Automatic curve sensor calibration method for an automotive CW/ICC system; Automatic curve sensor calibration method for an automotive CW/ICC system
• U.S. Pat. No. 5,031,330 Jul. 16, 1991; Title: Electronic Boresight; Inventors: Stuart; Kenneth (Los Altos Hills, Calif.)
• Assignee: Kaiser Aerospace & Electronics Corporation (Oakland, Calif.)

Other References

• Product description of Raytheon RT Secure, “Embedded Hard Real-Time Secure Operating System”, Copyright 2000, pp. 1-2.
• Product description of Raytheon RT Secure, Copyright 2001, pp. 1-2.
• Product description of Raytheon RT Secure, “Development Environment”, Copyright 2001, pp. 1-2.
• Product description of Raytheon Electronic Systems (ES), Copyright 2002, pp. 1-2.
• H. Chung, L. Ojeda, and J. Borenstein, “Sensor Fusion for Mobile Robot Dead-reckoning with a Precision-calibrated Fiber Optic Gyroscope”, 2001 IEEE International Conference on Robotics and Automation, Seoul, Korea, May 21-26, pp. 1-6.
• A. Das, R. Fierro, V. Kumar, J. Ostrowski, J. Spletzer, and C. Taylor, “A Framework for Vision Based Formation Control”, IEEE Transactions on Robotics and Automation, vol. XX, No. Y, 2001, pp. 1-13.
• J. Takezaki, N. Ueki, T. Minowa, H. Kondoh, “Support System for Safe Driving—A Step Toward ITS Autonomous Driving-”, Hitachi Review, vol. 49, No. 3, 2000, pp. 1-8.
• S. G. Goodridge, “Multimedia Sensor Fusion for Intelligent Camera Control and Human-Computer Interaction”, Dissertation submitted to the Graduate Faculty of North Carolina State University in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Electrical Engineering, Raleigh, N.C., 1997, pp. 1-5.
• M. Chantler, G. Russel, and R. Dunbar, “Probabilistic Sensor Fusion for Reliable Workspace Sensing”, pp. 1-14.
• ISIS Project: Sensor Fusion, Linkoping University Division of Automatic Control and Communication Systems in cooperation with SAAB (Dynamics and Aircraft), 18 pp.
• Hitachi Automated Highway System (AHS), Automotive Products, Hitachi, Ltd., Copyright 1994-2002, 8 pages.
• Vehicle Dynamics Lab, University of California, Berkeley, funded by BMW, current members: D. Caveney and B. Feldman, “Adaptive Cruise Control”, 17 pages.
• Counterair: The Cutting Edge, Ch. 2 “The Evolutionary Trajectory The Fighter Pilot-Here to Stay?” AF2025 v3c8-2, December 1996, pp. 1-7.
• Counterair: The Cutting Edge, Ch. 4 “The Virtual Trajectory Air Superiority without an “Air” Force?” AF2025 v3c8-4, December 1996, pp. 1-12.
• TNO FEL Annual Review 1998: Quality works, 16 pages.
• Boeing News Release, “Boeing Demonstrates JSF Avionics Multi-Sensor Fusion”, Seattle, Wash., May 9, 2000, pp. 1-2.
• Ada 95 Transition Support—Lessons Learned, Sections 3, 4, and 5, CACI, Inc.-Federal, Nov. 15, 1996, 14 pages.
• Joint Strike Fighter Terrain Database, ets-news.com “Simulator Solutions” 2002, 3 pages.
• MSRC Redacted Proposal, 3.0 Architecture Development, pages 1-43.
• Powerpoint Presentation by Robert Allen—Boeing Phantom Works entitled “Real-Time Embedded Avionics System Security and COTS Operating Systems”, Open Group Real-Time Forum, Jul. 18, 2001, 16 pages.
• Green Hills Software, Inc., “The AdaMULTI 2000 Integrated Development Environment”, Copyright 2002, 7 pages.

BACKGROUND

Next generation automotive systems such as Lane Departure Warning (LDW), Collision Avoidance (CA), Blind Spot Detection (BSD) or Adaptive Cruise Control (ACC) systems will require target information from multiple sensors including a new class of sensor such as radar, image or laser, similar to those found on advanced tactical fighter aircraft. For example, one sensor may be located on the front bumper of the vehicle and obtains range and azimuth information about vehicles and stationary objects in front of the vehicle. Another sensor may be located on the dash of the vehicle and obtains image information about vehicles and stationary objects in front of the vehicle. Another sensor may be located on the side of the vehicle and obtains either range and azimuth data or image data in order to determine velocity and track information on vehicles that pass the vehicle. These new systems must take all of the information from the multiple sensors on the vehicle and compute an accurate picture of the moving objects around the vehicle; this is known as kinematic state of the targets, or Situation Awareness (SA). To do this the Situation Awareness Platform (SAP) must accurately align the sensors to each other so that information about a target from one sensor can be used with information about the target from a different sensor. This is called Sensor Fusion (SF), this is necessary for the SAP to get an optimal kinematic state of the targets around the vehicle in order to assess threat. The sensors must also be aligned to the body of the vehicle so that the SAP can determine the position and velocity of the target with respect to the vehicle; this is called Navigation Fusion (NF).

One method of aligning the sensors to each other and to the vehicle is to use mechanical and optical instruments, such as auto-collimators and laser boresight tools, during the production of the vehicle. This technique is not only costly, but would be required if a sensor were repaired or replaced after production. An alignment procedure would have to be performed again in order to assure the safety critical systems were reporting accurately. Also as the vehicle goes through normal wear and tear the sensors would start to become misaligned and may not be noticed by the operator. This means that the data from the sensors would not correlate with each other and the vehicle reference frame until the sensors were aligned again. Again, this would be costly to the vehicle operator and until performed, the SAP may not provide accurate data. Therefore, a method to align the sensors to each other and to the vehicle without the use of sophisticated optical tools is required. This patent addresses this problem by describing methods that can be used to align the sensors to each other and to the vehicle that do not require external alignment equipment.

In a discussion of Prior Art, U.S. Pat. No. 5,245,909, Automatic Sensor Alignment, relates to systems for maintaining alignment-sensitive aircraft-borne avionics and weapons sensors in precise alignment. It further relates to methods for precisely aligning sensitive avionics for weapons system instrumentation, which is subject to vibrations causing misalignment. Whereas this disclosure relates to methods and systems that support advanced automotive systems not described in the prior art. A second key difference is the reliance on sensor data from the vehicle as part of the alignment method. Another difference is using image sensors with elements of the vehicle in the field of view of the imager and employing optical methods for determining changes to the alignment with respect to the vehicle and vehicle reference frame, then applying a compensation based on the misalignment angle measured. Finally, this system described herein does not require a reliance on boresighting and aligning any sensor to achieve a vehicle reference frame.

U.S. Pat. No. 6,202,027, Automatic Curve Sensor Calibration, describes an improved system for accurately determining the travel path of a host vehicle and the azimuth angle of a target vehicle through an automatic calibration that detects and compensates for misalignment and curve sensor drift. The difference is a reliance on observed objects and track file generation and subsequent changes to the track files over time. Whereas this patent teaches methods of alignment based force vectors, rotational rates or optically measured changes with respect to the vehicle reference frame. Essentially all observed objects are compensated for misalignment error on the observing vehicle.

U.S. Pat. No. 5,031,330, Electronic Boresight, teaches that pairs of level sensing devices can be used in a method that aligns plane surfaces to one another by tilting platforms equal to the amount misalignment measured to adjust the sensor azimuth. Whereas this patent teaches that the sensors are rigidly mounted to the vehicle and correction to misalignment is done by compensation values observed with respect to the vehicle reference frame.

SUMMARY OF THE INVENTION

A vehicle sensor system configured to gather sensory data 360 degrees around the vehicle, comprises sensors for gathering data such as: range (e.g. ultrasonic); range and azimuth (e.g. laser and/or radar); images (e.g. optical and/or thermal). The vehicle has sensors that align and establish a vehicle reference frame by measuring body yaw, pitch and roll rates as well as acceleration along the 3 axes of the vehicle. The imaging sensors that have a clear view of body mold lines, like hood or rear deck, will align themselves to the vehicle reference frame, those sensors that can not align using optical methods are aligned to the vehicle using accelerometers and rate sensors by reading the inertial acceleration or angular rotation to align themselves to each other. An Integrated Computing Platform (ICP) hosts the SAP software that maintains complete system alignment by determining differences in alignment and applying or updating a compensation value with respect to the vehicle body coordinates resulting in a dynamically boresighted system.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a diagram showing how a common inertial acceleration is sensed by accelerometers on each sensor and can be used to align the sensor coordinate frames.

FIG. 2 is a diagram showing the pitch angles used to determine the pitch misalignment angle of the optical sensor.

FIG. 3 is a diagram showing the yaw data that is used to determine the yaw misalignment angle of the optical sensor.

FIG. 4 is a diagram showing the roll data that is used to determine the roll misalignment angle of the optical sensor.

FIG. 5 is an image showing the top of the hood and how it is used to compute the pitch misalignment angle.

FIG. 6 is a magnified image of the hood line showing the pixels of the image.

FIG. 7 is an image showing the top of the hood and how it is used to compute the roll misalignment angle.

FIG. 8 is a magnified image of the banked hood line showing the pixels of the image.

FIG. 9 is an image showing the top of the hood and how it is used to compute the yaw misalignment angle.

FIG. 10 is a flow chart that shows the alignment process when all sensors have micro-inertials.

FIG. 11 is a flow chart that shows the alignment process when using micro-inertials and an optical sensor.

FIG. 12 is a flow chart that shows the alignment process when all of the sensors are optical.

FIG. 13 is a flow chart that shows the alignment when the sensors are on a common platform.

DETAILED DESCRIPTION

One method is to attach three axis accelerometers to each sensor and to the vehicle and use gravity and the acceleration of the vehicle, which will be sensed by the accelerometers, to align the sensor axes to each other and to the vehicle. Information from the vehicle that is available on the Car Area Network (CAN) bus will also be used to perform the calculation of the misalignment angles. FIG. 1 shows in two dimensions the relation between sensor A frame, sensor B frame and the vehicle body reference frame. There are two accelerometers that sense acceleration in the X and Y axes of the sensors and vehicle. This problem can easily be expanded to three dimensions with another accelerometer located in the Z-axes of each sensor and vehicle.

In FIG. 1 the vehicle experiences a linear acceleration and this common acceleration is observed by the accelerometers located on sensor A, sensor B and the vehicle body. The accelerometers that are attached to the vehicle body are aligned to the vehicle body reference frame. By taking the difference in acceleration data from the accelerometers on sensor A and sensor B and inputting this data in a Kalman Filter, the misalignment angle between the two sensors, θsa−θsb, can be computed. The same can be done between sensor A and the vehicle body, and sensor B and the vehicle body to compute all of the misalignment angles. This approach can be used to compute the three dimensional misalignment angles of roll, pitch and yaw between sensors and the vehicle body reference frame.

The same approach can be used when the vehicle is turning and each accelerometer group experiences a centripetal acceleration. However, in this case the difference in accelerations must be compensated by the centripetal acceleration resulting from the lever arm vector between the two sensors and the angular rotation of the vehicle. The angular rotation of the vehicle is sensed by a gyro triad or micro-inertial device located at the vehicle body reference frame
Acomp=Asensora−wxwxRl
The input to the Kalman filter is now:
Acomp−Asensorb
where:

• • Asensora is the acceleration measured by sensor A accelerometers
  • Asensorb is the acceleration measured by sensor B accelerometers
  • w is the angular rotation of the vehicle measured by the ref gyros
  • x is the cross product of two vectors
  • Rl is the lever arm vector between sensor A and sensor B
  • Acomp is the sensor acceleration compensated for lever arm rotation

Also if the vehicle is stationary, the accelerometer groups will sense gravity and this can be used to help compute some of the misalignment angles. Information from the vehicle CAN bus, such as wheel rotation speeds are zero, will tell the Kalman filter that the vehicle is not moving and the only sensed acceleration will be from gravity.

FIG. 10 is a flow chart showing the process when all of the sensors, as well as the vehicle body, have a micro-inertial attached to them. When the vehicle is moving, the micro-inertials sense the angular rotation and/or acceleration of the vehicle and this information is the input to a Kalman filter. The filter uses this information to estimate the roll, pitch and yaw misalignment angles between a sensor and the vehicle body frame. These misalignment angles are then used to rotate the sensor target data into the vehicle body frame. With all of the target data in a common reference frame the processor can fuse data from several sensors into an optimal target track file.

The second method is to use accelerometers to align the sensors to each other and one of the sensors is aligned to the vehicle body by using optical information from the sensor itself. For example, acceleration data can be used to align sensor A to sensor B, but sensor B is aligned to the vehicle body directly by using sensor B to compute the misalignment angles between sensor B and the vehicle body. Since sensor A is aligned to sensor B and sensor B is aligned to the vehicle body, you can compute the misalignment between sensor A and the vehicle body. Sensor B can be a visual sensor, such as a video camera, and by observing the outline of the hood and body of the vehicle using this camera, you can compute the misalignment angles between sensor B and the vehicle body frame.

FIG. 2 shows that the pitch misalignment angle is the angle between the sensor's X-axis and vehicle's X-axis in the vertical plane. The pitch angle between the vehicle X-axis and a line from the sensor to the top point of the hood, Φvehicle, can be computed from the vehicle's dimensions. The image from the sensor, FIG. 5 for example, shows the top of the hood. By counting the pixels from the center of the image down to the hood, Pp, the sensor pitch angle can be computed. Using a 480×640 pixel image, this angle can be computed to within 1 pixel, see FIG. 6. With a vertical field of view, FOVv, the pitch angle is:
Φs=(Pp/480)*FOVv
The pitch misalignment angle is:
Φmisalign=Φs−Φvehicle

FIG. 3 shows that the small yaw misalignment angle is the angle between the sensor's X axis and vehicle's X axis in the horizontal plane. The sensor image shows the left and right edges of the hood, FIG. 9. By computing the pixels from the left hood edge or mark on the hood to the left of the image border, Pyl, and the right hood edge or mark to the right border, Pyr, the yaw angle of the sensor misalignment with a horizontal field of view, FOVh is:
ΨMisalign=((Pyl−Pyr)/2*640)*FOVh

FIG. 4 shows that the small roll misalignment angle is the angle between the sensor's Y-axis and vehicle's Y-axis in the vertical plane. The sensor image shows that the hood line and the sensor level lines cross to form the roll misalignment angle. This is shown in FIG. 7. By measuring the pixels between the two lines at the edge of the image, Pr, the roll misalignment angle can be computed as follows:
ΘMisalign=(2*Pr/640)*180/π
FIG. 8 shows that the hood line can be determined accurately to within a couple of pixels.

FIG. 11 is a flow chart showing the process when at least one of the sensors is an optical device. All of the sensors have a micro-inertial attached to them. The optical device can see the targets and the outline of features of the vehicle, such as the hood line. The optical sensor uses the hood line information to compute the roll, pitch and yaw misalignment angles between the optical sensor frame and the vehicle body frame.

When the vehicle is moving, the micro-inertials sense the angular rotation and/or acceleration of the vehicle. Like FIG. 10, the Kalman filter estimates the roll, pitch and yaw misalignment angles between a sensor frame and the optical sensor frame. These misalignment angles as well as the misalignment angles between the optical sensor and the vehicle body frame are then used to rotate all of the sensor target data into the vehicle body frame. Again, with all of the target data in a common reference frame the processor can fuse data from several sensors into an optimal target track file.

A third method is to use optical information from sensor A and sensor B to compute the misalignment between the two sensors and to use optical information from sensor B to compute the misalignment between sensor B and the vehicle body. For example, sensor A can be a ranging laser sensor and it sends out multiply beams of light to detect a target. When the light is reflected from the target, sensor B can also detect the reflected light in its video camera and using this information it can compute the misalignment between sensor A and sensor B.

FIG. 12 is a flow chart showing the process when all of the sensors on the vehicle are optical sensors. Each optical device can see targets and the outline of features of the vehicle, such as the hood or truck line. The optical sensors use this vehicle body information to compute the roll, pitch and yaw misalignment angles between the optical sensor frame and the vehicle body frame. These misalignment angles are then used to rotate the sensor target data from each sensor into the vehicle body frame. Like the two cases above, with all of the target data in a common reference frame the processor can fuse data from several sensors into an optimal target track file.

A fourth method is to collocate all of the sensors into one box that is mounted on the vehicle, such as the roof, so that all sensors are always aligned with respect to each other and the only alignment required is the alignment between this sensor box and the vehicle body. This can be performed by using a set of accelerometers in the sensor box and on the vehicle body frame or optically by using a video camera in the sensor box.

FIG. 13 shows the case where all of the sensors are mounted onto one fixed platform. If one of the sensors is an optical sensor then it can be used to align the platform frame to the vehicle body frame as shown above. Once this set of misalignment angles is computed, then all of the target data from all of the sensors can be rotated to the common vehicle body reference frame. As shown above all of the target data is now in one reference frame for computing the optimal target tracks. If none of the sensors are optical, then a set of micro-inertials can be mounted on the common platform and also on the vehicle body. While the vehicle is moving the Kalman filter can now be used to compute the misalignment angles as discussed in the above paragraphs.

The systems described above can use dedicated processor systems, micro controllers, programmable logic devices, or microprocessors that perform some or all of the operations. Some of the operations described above may be implemented in software and other operations may be implemented in hardware.

For the sake of convenience, the operations are described as various interconnected functional blocks or distinct software modules. This is not necessary, however, and there may be cases where these functional blocks or modules are equivalently aggregated into a single logic device, program or operation with unclear boundaries. In any event, the functional blocks and software modules or described features can be implemented by themselves, or in combination with other operations in either hardware or software.

Having described and illustrated the principles of the invention in a preferred embodiment thereof, it should be apparent that the invention may be modified in arrangement and detail without departing from such principles. Claim is made to all modifications and variation coming within the spirit and scope of the following claims.