Classifications

• • G—PHYSICS
  • G01—MEASURING; TESTING
  • G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
  • G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
  • G01C21/10—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration
  • G01C21/12—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning
  • G01C21/16—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning by integrating acceleration or speed, i.e. inertial navigation
  • G01C21/183—Compensation of inertial measurements, e.g. for temperature effects
  • G01C21/188—Compensation of inertial measurements, e.g. for temperature effects for accumulated errors, e.g. by coupling inertial systems with absolute positioning systems

Definitions

• This invention relates to tracking and control of tiltable bodies. It is known to use quaternions to represent the orientation in space of an object.
• Quaternion notation is, in general, more computationally efficient to use than the more widely employed Euler data representation. Furthermore, quaternion notation is not subject to singularities which can occur when using Euler notation.
• the following U.S. patents disclose using quaternions to control, determine and/or display the orientation of an object in space: Nos. 5,875,993; 5,212,480; 4,797,836; 4,742,356; and 4,737,794.
• a quaternion is a four-element, hypercomplex number first conceived by Sir William Rowan Hamilton in 1843.
• a quaternion consists of a scalar part and a complex vector part.
• the vector part consists of an ordered triple (vector) of three real components which are assigned direction by three orthogonal complex unit vectors: i, j, k.
• An example of a general quaternion Q is shown below:
• the square magnitude of the quaternion can be computed by forming the product of the quaternion with its complex conjugate as shown:
• a vector rotation is accomplished using a one-sided rotation operator, R, which in three-dimensional space can be represented as a real, 3x3 orthogonal matrix.
• the resulting vector, X' is rotated about a general axis by a specific angle, both of which are determined by the unit quaternion, Q. If the axis of rotation is denoted by a unit vector, n,
• the quaternion components thus defined are also called the Euler Parameters. These parameters contain all of the information necessary to derive the axis and angle of rotation.
• the rotation axis defined by unit vector, n, is also called the eigenaxis since it is the eigenvector of
• the axis of rotation must be common to both the original and rotated frames and therefore must be unchanged by the rotation operator.
• the so-called eigenaxis rotation is a single rotation about a general axis as compared to the Euler angle rotation which accomplishes the same transformation by performing three separate rotations: yaw, pitch, and roll, about the z, y, and x axes, respectively.
• the quaternion complex unit vectors (i, j, k) are related to the Pauli spin matrices as shown:
• the unit quaternion can be written as:
• This matrix form has many useful properties. For example, it can be shown that the reflection of the vector, x, through a plane defined by the unit normal, a, is easily produced using the matrix form of the vectors as shown:
• vectors, a and b must be perpendicular to the axis of rotation and separated by an angle of ⁇ /2.
• the quaternion and the associated rotation matrix responsible for the transformation can be uniquely determined.
• the attitude of the system can be uniquely determined by taking additional "sightings" of the same two external reference points from the rotated frame and comparing the coordinates of the reference vectors measured in the final frame to those measured in the initial frame.
• quaternion components may also be contemplated to compute quaternion components by direct integration of a quaternion rate. This requires an expression for the derivative of the quaternion as a function of angular rate. This can be derived by differentiating Eq. 21. The resulting derivatives of the quaternion coefficients are each a linear combination of the angular rate components weighted by the existing quaternion components. If the quaternion and angular rates are placed in vector form, the following matrix equation results:
• the orientation and/or rotation of a body may be represented in more than one frame of reference.
• a frame of reference relative to the body of interest itself, or to define a frame of reference relative to an external fixed object.
• the earth may be taken as a fixed object for defining a frame of reference.
• Eq. 28 the angular rate is referenced to the body frame.
• a similar matrix equation in which the angular rate is referenced to the earth frame is shown below as Eq. 29. Note that shifting the reference frame produces several sign changes in the quaternion matrix.
• the quaternion rate equation When the quaternion rate equation is integrated using the instantaneous angular rate of the body in question, the resulting quaternion contains the information necessary to transform coordinates from the initial frame of reference to the rotated frame.
• the quaternion components can be used to form the rotation matrix, R, described previously (Eq. 8).
• the rotation matrix which transforms body coordinates into earth coordinates is shown below in terms of quaternion components:
• R EB 2 q 2 a i + ⁇ 3 ⁇ o ) 00 - 01 2 + 02 2 - 03 2 2(0203 - 0100 ) 2 2
• the Euler method depends heavily on trigonometric functions and is quite computationally intensive.
• the Euler rate equations depend on the secant of the pitch angle which results in a singularity when the pitch reaches +90 degrees.
• the quaternion method contains no trigonometric functions and relies only on multiplication and addition. Consequently, the quaternion method is far more computationally efficient.
• the attitude of a body can be uniquely represented in quaternion form, given two vectors defined in the body frame and two external reference points. However, it can be difficult and/or expensive to establish two external reference points, particularly in the case of terrestrial objects, whether located on land or on or under the surface of a body of water.
• the present inventor believes he is the first to recognize how quaternion notation may be usefully modified such that only one external reference point is needed to generate the modified quaternion form, and that such a modified quaternion form may be usefully applied to bodies which are free to tilt, but for which the heading (yaw) is constrained or need not be controlled by the system which employs the modified quaternions.
• An aspect of the invention provides a method of estimating the attitude of a tiltable body, where the body includes tilt sensing devices such as a fluidic tilt sensors and angular rate sensing devices such as gyros.
• the method includes the steps of: outputting angular rate information from the angular rate sensing devices, transforming and integrating the outputted angular rate information to produce first quaternion position information such that the first quaternion position information is constrained to represent a rotation about a horizontal axis in an earth frame of reference, outputting tilt information from the tilt sensing devices, processing the outputted tilt information to generate second quaternion position information such that the second quaternion position information is constrained to represent a rotation about a horizontal axis in the earth frame of reference, comparing the first quaternion position information to generate error information, and using the error information to compensate for drift in the angular rate sensing devices.
• Another aspect of the invention provides a method of estimating the attitude of a tiltable body, where the body includes tilt sensing devices such as fluidic tilt sensors and angular rate sensing devices such as gyros.
• tilt information outputted from a fluidic sensor or the like is used to compensate for drift or other offsets in a high- bandwidth angular rate sensor, such as a set of gyros.
• the error compensation provided through the tilt reference data allows the corrected high-bandwidth angular rate data to be integrated to provide a real-time attitude tracking estimate that does not accumulate offset errors.
• Heading angular rate information (yaw rate information) is discarded to produce a modified (3- component) quaternion representation that is compared with a similar modified quaternion rate information derived from tilt information and based on a single, gravitational reference point. Error compensation and filtering are performed with quaternion notation to provide computational efficiency.
• Fig. 1 is a schematic and generalized side view of a tiltable body with respect to which the present invention is applied.
• Fig. 2 is a top plan view of the tiltable body of Fig. 1.
• Fig. 3 is a block diagram illustration of a control system for the tiltable body of Fig. 1.
• Fig. 4 is a block diagram representation of an attitude and rotation estimation portion, provided in accordance with the invention, of the control system of Fig. 3.
• Fig. 1 is a schematic side view of a generalized tiltable body 10 to which the present invention is applicable.
• Fig. 2 is a top plan view of the same body 10.
• the body 10 may be any object for which inclination from a level condition is to be tracked, represented and/or controlled.
• the body 10 may be a robot, a tiltable motor-controlled wheelchair, an off-shore drilling platform, a surface ship or boat or submarine water craft, a tiltable railway locomotive or car, a passenger-bearing enclosure or vehicle in an amusement ride, a flight simulator, or a vessel used to contain and selectively pour out a material used in a chemical reaction or other manufacturing process.
• Body 10 may also be an automobile.
• a control system 12 which receives input from one or more tilt sensing devices 14 and one or more angular rate sensing devices 16.
• the tilt sensing devices 14 may be conventional fluidic tilt sensors and/or accelerometers. It is contemplated that there be one or more tilt sensors.
• the angular rate sensor 16 may be constituted by conventional gyros or other known angular rate sensing devices. It is contemplated that there be one or more angular rate sensors. A coordinate system for the ensuing discussion may be established as illustrated in Figs.
• the horizontal forward direction is considered to be the positive x axis as indicated by arrow 20.
• the positive y axis is considered to be in the horizontal direction pointing to the right and perpendicular to the x axis, as indicated by arrow 22 (Fig. 2).
• the positive z axis is considered to be pointing vertically straight down, as indicated by arrow 24 (Fig. 1).
• Tilt will be defined as an angular deviation from vertical, having a range limited to plus or minus 180°.
• Pitch is defined as rotation around the y axis with positive pitch being a backward tilt.
• Yaw is defined as rotation around the z (vertical) axis, with positive yaw being defined as rotation to the right. It will be noted that these definitions adhere to a right-hand rule.
• the above definitions of pitch and roll are similar to those for Euler pitch and roll, except that Euler angles are applied in a specific order and produce intermediate frames of reference. In the tilt-based system discussed herein, it is assumed that rotations around the vertical axis are nonexistent or can be disregarded.
• the heading of the body 10 may be fixed, or irrelevant to tilt tracking and control, or subject to control by a human operator, or controlled by a system that does not or need not take tilt information into account.
• Fig. 3 illustrates, in block diagram form, components of the control system 12 shown in Fig. 1.
• the control system 12 includes an interface 40 through which the control system 12 receives signals outputted from the tilt sensors 14 and gyros 16. These signals are processed by an attitude/rotation estimation block 42, in accordance with the invention.
• block 42 On the basis of the signals outputted from the gyros and the sensors, block 42 provides an estimate of the attitude and/or angular rotation rates of the body 10, in a manner which will be described below.
• a block 44 On the basis of the attitude and/or rotation information provided by block 42, a block 44 generates control signals for controlling actuators 46 to control the attitude of the body 10.
• the actuators may include motors, solenoids, floodable chambers, or other conventional mechanical, electromechanical, hydraulic or pneumatic devices used to control the attitude of the types of bodies referred to above.
• Electronic hardware for carrying out at least part of blocks 42 and 44 may be constituted by one or more conventional microprocessors.
• the control signal generation block 44 also preferably includes driving circuitry for providing suitable driving signals for the actuators 46.
• control system 12 is indicated in Fig. 1 as being carried on the body 10, it should be understood that at least some portions of the control system 10 may be physically separate from the body 10.
• processing circuitry to carry out some or all of the blocks 42 and 44 may be remote from the body 10 and may receive output from the tilt sensors 14 and gyros 16 via telemetry, and may also transmit control signals back to the body 10, by suitable wireless communication channels.
• Tilt sensors such as conventional fluidic sensors, may be relatively free of drift and therefore provide a reliable indication of a gravity vector.
• attitude of a static system can be determined relative to gravity using such sensors.
• sensors of this type are sensitive to angular and vibrational accelerations, and therefore must be low-pass filtered to attenuate the effects of accelerations other than gravity.
• the resulting data provides the average direction of gravity over a finite period of time determined by the bandwidth of the filter.
• gravitation sensors alone are not adequate.
• gyros or other angular rate sensors may provide high bandwidth and rapid response, without being adversely affected by acceleration.
• the rate data can be integrated to produce position information.
• rate sensors are subject to drift which can cause significant errors when integrated.
• the estimation process provided by block 42 combines the high and low bandwidth information respectively provided by the gyros 16 and the tilt sensors 14 to provide accurate position and rotation data, using a model-based estimator topology.
• a model-based estimator an estimate of the true position is produced by comparing sensor data with an internal model of the system dynamics. The error between the measured data and the predicted data is used to continuously refine the estimate. The degree to which this error affects the estimate is determined by a weighting matrix, H, that feeds the appropriate state errors back to the estimator.
• H weighting matrix
• the internal system model is defined by matrices (A,B,C,D).
• the feedback matrix H determines the stability of the estimator.
• the estimate is a frequency-dependent weighted sum of the two measurements.
• the Laplace transform of this result is shown below:
• the estimated position approaches the low bandwidth measurement, ⁇ 0 , as before.
• the estimated position approaches the integral of the high bandwidth rate
• this type of estimator is well suited for producing an estimate of angular position and rate from low bandwidth tilt data and high bandwidth angular rate data.
• the gains ki and k 2 are chosen such that the appropriate filter bandwidth and stability
• quaternions can be used to track the attitude of a system through any angular motion. Moreover, quaternions are a computationally efficient notation and also are immune to the singularities encountered by Euler angles. It is also known that data from high and low bandwidth sensors can be combined using an estimator to provide accurate data over the full bandwidth. However, quaternions do not lend themselves to direct physical measurement, and this complicates the derivation of an absolute quaternion reference from tilt sensor data. Without such a reference, it is impossible for the estimator to prevent gyro drift from corrupting the integrated quaternion and eventually causing the system to lose track of its angular position.
• the three angular rate components and the three modified quaternion components all contain necessary information.
• the yaw component of the angular rate is not important because the system is a tilt-based. Therefore, the true yaw component can be replaced with a virtual yaw component that satisfies the above constraint. In practice, this may be done by transforming the measured angular rate from the body frame into the earth frame. The resulting earth yaw can then be replaced with a virtual earth yaw which satisfies the following constraint equation:
• the purpose of the estimator is to compare the quaternion obtained by integration of the quaternion rate to the quaternion obtained from the tilt data. Assuming that any steady-state error is due to gyro drift only, the magnitude of this error can be used to adjust the gyro drift correction term.
• the estimator gains are chosen such that the steady-state error is, over time, driven to zero. Selection of estimator gain values is well within the abilities of those who are skilled in the art, and therefore need not be discussed further.
• the quaternion error In order to properly adjust the correction term, the quaternion error must be converted back to an angular rate error. This is accomplished using the original quaternion rate equation which has been solved for angular rate in the earth frame as shown:
• the tilt data is converted to a modified quaternion. It has been shown that the quaternion components can be produced by forming the dot and cross products of two unit vectors which are perpendicular to the axis of rotation and
• the desired unit vectors can be constructed as follows:
• the quaternion components can be generated by computing the dot and cross products as shown below:
• the cross product of a and b produces a vector which is perpendicular to the gravity vector in each frame. Consequently, this vector must also be horizontal. This implies that the z-component of the quaternion is identically zero as desired. Therefore, only the first three elements of the quaternion are non-trivial and satisfy the form of the modified quaternion.
• the estimator topology is split into two loops: the ki loop and the k loop.
• the kj loop provides low bandwidth filtering of the data from the tilt sensor, and determines the amount of error that is used to adjust the estimated quaternion rate.
• the k loop provides drift correction for the gyro data. Since the gyros are in the body frame, the k 2 loop must lie on the body side of the coordinate transformation.
• the k gain determines the amount of error that is used to correct the drift in the gyro data.
• the weighted error is then integrated so that the loop will, over time, achieve zero steady-state error. In other words, when the quaternion error reaches zero, the output of the k 2 integrator stops changing and the drift correction term remains constant.
• a summation block 50 in the k loop applies a weighted error signal (which may also be considered a drift correction signal) to the angular rate information outputted from the gyros 16, to produce corrected angular rate information.
• the corrected rate information is converted to the earth frame of reference and transformed into a modified quaternion at block 52.
• the calculations applied in that block are in accordance with equations 43 and 42, which were discussed above.
• the resulting estimated rate information which is in the above-described modified quaternion form is provided as an input to a summation block 54.
• a correction provided by the ki loop is applied to the estimated quaternion rate information to produce data indicative of the change in the estimated position.
• This data (which may be referred to as an estimated position difference signal) is, in turn, integrated at block 56 to provide the modified quaternion estimated position data.
• the modified-quaternion estimated position data outputted from block 56, and also the estimated modified-quaternion rate information outputted from block 52 are converted to Euler angles to be outputted to the control signal generation block 44 (Fig. 3). This conversion is easily accomplished using the following equations. (Although these equations use trigonometric functions, the computational complexity is minimal and can easily be performed using lookup tables.)
• control signal generator 44 is of the type which outputs quaternion signals rather than Euler signals
• block 58 may be omitted and the estimated quaternion position and rate information (in the modified form provided in accordance with the invention) may be provided directly to the control signal generator.
• the estimated position information in modified quaternion form is provided as an input to a summation block 60, at which it is compared to the current tilt information outputted from the tilt sensors 14, as converted to the modified quaternion form at block 62.
• the signal produced at block 62 may be referred to as a modified quaternion reference position signal.
• the conversion carried out at block 62 is in accordance with the procedure described above in connection with Eqs. 46-48.
• the resulting error signal outputted from summation block 60 is weighted by the gain factor ki at block 64, and then is applied as the above-mentioned correction signal at block 54.
• the output from block 60 (which may be considered an error position signal) is also provided as an input to block 66.
• block 66 converts the error between the estimated quaternion and the tilt quaternion into a tilt angle error.
• the invalid yaw component of the resulting signal output from block 66 is replaced with the earth- frame yaw component which is available from the processing of block 52, and the resulting angular error vector is converted to the body frame of reference at block 68.
• the output of block 68 is then integrated and weighted by the k 2 gain at block 70, and the resulting signal is applied at summation block 50 as the above-mentioned drift correction signal.
• the coefficients required for the processes in blocks 52, 66 and 68 are taken from the modified-quaternion position information outputted from the integration block 56, as is indicated at 72 in Fig. 4.
• attitude and rotation estimator may advantageously be performed by one or more suitably programmed computing devices (e.g. one or more microprocessors) which constitute part of the control system 12.
• computing devices e.g. one or more microprocessors
• attitude and rotation estimator provided in accordance with the invention, high bandwidth information provided by rotation sensors and low bandwidth information from tilt sensors are combined in a novel modified quaternion notation having three elements instead of the conventional four elements.
• the resulting estimated quaternion is valid over the designed bandwidth and does not drift over time.
• the computations to be performed can be handled efficiently using simple arithmetic such as multiplication, addition and square root.
• the modified quaternion estimator is applicable to a broad range of systems in which tilting is to be tracked and/or controlled.

Priority Applications (10)

Applications Claiming Priority (2)

Publications (2)

Family

ID=23974120

Family Applications (1)

Country Status (17)

Cited By (6)

Families Citing this family (82)

Citations (1)

Family Cites Families (8)

• 2000
  • 2000-02-03 US US09/496,785 patent/US6377906B1/en not_active Expired - Lifetime
• 2001
  • 2001-02-01 JP JP2001556277A patent/JP4808354B2/ja not_active Expired - Lifetime
  • 2001-02-01 AT AT01906867T patent/ATE329228T1/de active
  • 2001-02-01 RU RU2002120799/28A patent/RU2253092C2/ru active
  • 2001-02-01 ES ES01906867T patent/ES2266158T3/es not_active Expired - Lifetime
  • 2001-02-01 DE DE60120363T patent/DE60120363T2/de not_active Expired - Lifetime
  • 2001-02-01 BR BRPI0108066-0A patent/BR0108066B1/pt active IP Right Grant
  • 2001-02-01 CA CA002398704A patent/CA2398704C/en not_active Expired - Lifetime
  • 2001-02-01 MX MXPA02007560A patent/MXPA02007560A/es active IP Right Grant
  • 2001-02-01 WO PCT/US2001/003290 patent/WO2001057474A1/en active IP Right Grant
  • 2001-02-01 AU AU34727/01A patent/AU777647B2/en not_active Expired
  • 2001-02-01 PT PT01906867T patent/PT1257784E/pt unknown
  • 2001-02-01 EP EP01906867A patent/EP1257784B1/en not_active Expired - Lifetime
  • 2001-02-01 IL IL15100601A patent/IL151006A0/xx active IP Right Grant
  • 2001-02-01 CN CN01807823.0A patent/CN1270162C/zh not_active Expired - Lifetime
  • 2001-02-01 DK DK01906867T patent/DK1257784T3/da active
• 2002
  • 2002-07-31 NO NO20023641A patent/NO335201B1/no not_active IP Right Cessation
  • 2002-07-31 IL IL151006A patent/IL151006A/en unknown

Patent Citations (1)

Also Published As

Similar Documents

Legal Events