JP2013024845A - Tightly coupled gps and dead reckoning navigation for vehicle - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide tightly coupled GPS and a dead reckoning navigation system which collect data of a wheel speed transducer on a vehicle network and calculate a vehicle range and a vehicle heading direction.SOLUTION: Dead reckoning navigation fills a gap in navigation solution that occurs when GPS signal transmission is lost in general situations such as a tunnel, a parking lot, and the like. GPS fixes calculate continuous calibration for wheel radii and correction for speed effect, and the calculation improves performance and accuracy in the dead reckoning navigation during a long suspension period of GPS signal reception. When the GPS signal reception is recovered, the dead reckoning navigation solution provides a high quality start point for a GPS receiver for searching environment.

Description

The present invention relates to car navigation systems, and more particularly to road map correction feedback for a tightly coupled combination of a global positioning system (GPS) receiver and dead reckoning (DR) in a vehicle.

  GPS navigation systems with street map displays are now widely available in most new vehicles around the world. New map updates appear to be purchased and installed on DVD discs and published annually on a yearly basis. Typical GPS navigation receivers used in these vehicles are tens of meters accurate in good environments, but can drop to 100 meters in dense urban environments. These car navigation systems assume that the vehicle is traveling on a road on the map and therefore use the map information to snap the vehicle location icon to the exact route of the nearest road, but GPS reception If lost in the same way in tunnels and parking lots, the user position on the map display screen is simply not updated.

  Factory installed navigation systems are generally expected to provide higher reliability and accuracy compared to portable navigation devices in the repair parts market. The driver will not tolerate long start-up times or navigation failures simply due to entering a tunnel or parking lot. Thus, conventional factory-installed navigation systems include inertial sensors such as speedometers, travel recorders, gyros and accelerometers, and vehicle data buses for accessing speed, forward / reverse, and mileage information. It is connected. However, inertial sensors impose higher costs.

  Obtaining arbitrary course information in conventional systems requires that an expensive and delicate directly coupled automotive gyroscope be installed. SiRF Technology, Inc. (San Jose, Calif.) Has sold a solution that they say will eliminate this relatively expensive sensor and instead use sensor data already available from other vehicle subsystems. ing. According to SiRF, sensor data measurements must be accurate enough to extract meaningful information for dead reckoning, and sufficient frequency must be provided for sensor measurements. Also, any delay in sensor data delivery to the dead reckoning system must be short, for example, a latency of 10 milliseconds or less.

  Most new vehicles in North America and Europe are equipped with standard vehicle speed sensors and anti-lock braking systems (ABS). Compass is more frequently seen in the United States than in Europe, but stability control is more common in Europe than in the United States. Such dead reckoning software must take into account the various availability of a particular sensor. The early SiRFdrive 1.0 was configured for a vehicle platform with a direct-coupled gyroscope, while the later SiRFdrive 2.0 used distributed ABS module sensor data instead. SiRFdrive 2.0 calculates the individual wheel speed and determines the speed of the vehicle 110 along with its rotational speed, so that an expensive gyroscope can be eliminated. The SiRFdrive 2.0 dead reckoning system is described in the press release as collecting wheel rotation ticks (ie speed pulses) from each wheel. It further determines the calibration value required for accurate dead reckoning. The better the resolution of the wheel increment data (for example, the greater the number of wheel pulses per revolution), the better the overall dead reckoning performance.

  Since the wheel rotation step data is read directly from the vehicle bus, there was no need to connect to the wheel sensor itself. However, using raw ABS data is not so easy. Dead reckoning requires the highest resolution at low speed because the maximum course change usually occurs when turning a road intersection at low speed. The main function of the ABS module is to control the individual wheel locks that occur when a strong braking system is applied at high speeds. Thus, some ABS modules do not send measurements at their maximum resolution, and wheel sensor ABS modules may output inaccurate data at low speeds or no data at all. In these cases, the ABS software is usually upgraded and can provide the necessary solution and the low-speed data output required by dead reckoning applications. The advertising literature says that SiRFdrive 2.0 has sophisticated algorithms and such data loss minimizes the negative effects of having accuracy at low vehicle speeds.

  Unfortunately, such a wheel-driven dead reckoning system does not meet those promises. What is still needed is a practical system with self-calibration and correction algorithms that work in combination with GPS and dead reckoning.

  Briefly, the tightly coupled GPS and dead reckoning system of the present invention, in one embodiment, collects wheel speed transducer data on the vehicle network and computes vehicle range and direction. Dead reckoning bridges the gap in navigation solutions that occur when GPS signal transmission is lost in tunnels, parking lots, and other common situations. Continuous calibration of wheel radius and compensation for speed effects is computed from the GPS fix, which improves dead reckoning performance and accuracy during long outages of GPS signal reception. When the GPS signal is recovered, this dead reckoning solution provides a high-quality starting point for GPS receivers that search the surroundings.

  A car navigation system according to the present invention is mounted on a vehicle and receives a GPS signal transmission obstacle from an orbiting navigation satellite, and is coupled to the GPS receiver, and dead reckoning (DR) during the signal fault period. A DR computer that provides range and course information computed by the system, wherein vehicle navigation solutions and assistance are continuously available through a screen display to the user, coupled to the DR computer, A car navigation system comprising a wheel tick sensor for providing rotation data relating to the rotation of individual supporting wheels, characterized by an estimate of the true radius of said individual wheel, and an estimation of the true radius of said individual wheel Values to generate information about range changes (ie delta range) and course changes (ie delta course), the behavior of the vehicle Combined with the rotation data from the wheel tick sensor, the delta range and delta course information is propagated by the DR computer from absolute position information periodically obtained from the GPS receiver, and the dead reckoning navigation result is Alone, possibly as a vehicle navigation solution, and the vehicle navigation solution and assistance are continuously available through a screen display to the user, alone from the GPS receiver, and alone by the DR computer. Or derived from a composite provided by the GPS receiver and the DR computer.

  The car navigation system of the present invention selects the vehicle navigation solution from the GPS receiver alone, from a DR computer alone, or from a composite provided by both the GPS receiver and the DR computer, A computer program for deriving is further provided.

  The car navigation system of the present invention includes a computer program that operates in one of three modes including a GPS receiver alone, a DR computer alone, and a combined mode that includes both the GPS receiver and the DR computer. In addition, the operation in the GPS receiver single mode occurs when a GPS fix is available, there is a satellite vehicle available sufficiently consistently, and more than two satellite vehicles are being tracked. Operation in the DR computer single mode occurs when GPS fix is not available, or when two or less satellite vehicles are being tracked and calibration is available, and operation in the combined mode No more than two satellite vehicles have been tracked from inspection or are available consistently Vehicle absence, and calibration are available, characterized in that occur when it is confirmed.

  The car navigation system of the present invention comprises a programmable permanent stop bit and a conditional program branch that immediately follows activation of the GPS measurement hardware, the conditional branch being the programmable permanent stop bit. Transfer control to the DR computer-only mode when calibration is available and the programmable persistent stop bit is true, or the programmable persistent And further comprising transferring control to the GPS only mode when a GPS fix is available if the target stop bit is chaotic.

  The car navigation system of the present invention further comprises an automatic process for continuous calibration of an estimate for the true radius of each of the individual wheels, and absolute position information obtained periodically from the GPS receiver Is compared with the rotation data provided by the wheel tick sensor.

  The car navigation system of the present invention has a buffer and time stamp processing for receiving a packet of wheel step rotation data received from a controller area network (CAN), and estimates the wheel step rotation data from the buffer and time stamp processing. And a dead reckoning Kalman filter for calibration.

  The car navigation system of the present invention is computed to approximate the true radius of each of the individual wheels and is used to provide an estimate of the delta range, delta course, and delta course total for dead reckoning solutions The initialization value is further provided.

  The car navigation system of the present invention further includes speed effect compensation for adjusting an estimated value of a wheel radius.

The application program of the present invention is implemented as a GPS application on a host computer and is coupled to receive measurements from a GPS measurement platform (MP) on a vehicle with wheels, tightly coupled GPS, dead reckoning navigation, And a map matching navigation software application program that collects time, fix trigger, Doppler, and pseudo-range reports from GPS measurement platform (MP) devices mounted on wheels-equipped vehicles; Asynchronous data generated by the vehicle is aligned to GPS time without this processing, offset calculation processing from GPS time to host computer time, and position based on each new position fix and GPS system time (P ), Speed (V), wheel radius and ratio calibration • Parameter (B), Delta course and delta range (D), and least square snapshot estimator over time and calibration parameters (B), delta for position (P), speed (V), wheel radius and ratio To the availability of GPS solutions and wheel data calibration, as well as Kalman filters configured for heading and delta range (D), and time to handle dead reckoning delta range, delta heading, compound mode, and propagation Depending on the GPS alone mode, dead reckoning alone mode, or continuous mode selection process that operates in both GPS and dead reckoning combined mode, the hardware pre-data calculation and GPS connected to the time offset calculation process New fix calculation to adjust millisecond timer and measurement by vehicle wheel transducer A buffer process of collecting CANbus data time stamps relating to wheel rotation increment which, A _L and A _R calibration parameters for delta course / tick difference between the left and right wheels, as well as, for DeltaRange / tick of the left and right wheels B _L and B _R calibration parameters, as well as dead reckoning calibration Kalman filters that process the head offset calibration parameters that translate the total delta course to a known course reference, followed by the Kalman filter Application programming to convert the data from the buffer processing into delta range, delta heading, and delta heading sum information, including dead reckoning error model, and output new navigation fix・ Interface (API Characterized in that it comprises and.

  The application program of the present invention includes a mode selection program that operates in one of three modes including a GPS receiver alone, a DR computer alone, and a combined mode that includes both the GPS receiver and the DR computer. In addition, the operation in the GPS receiver single mode occurs when a GPS fix is available, there is a satellite vehicle that is available sufficiently consistently, and more than two satellite vehicles are being tracked. Operation in the DR computer single mode occurs when GPS fix is not available, or when two or less satellite vehicles are being tracked and calibration is available, and operation in the combined mode No more than two satellite vehicles have been tracked from inspection, or are sufficiently consistently available Kuru is no, and a calibration is possible, occur when it is confirmed, the new navigation fix is characterized in that it is then output through the API.

1 is a functional block diagram of an embodiment of a tightly coupled GPS and dead reckoning navigation system of the present invention for a four wheel vehicle with a vehicle bus network. FIG. FIG. 2 is a functional block diagram of two possible implementations of the combination of the GPS receiver of FIG. 1 and dead reckoning (DR) computer. Functions of the GPS and DR combined embodiment of the present invention that can be implemented with various hardware and software, such as in FIGS. 2 (a) and (b), for the system application shown in FIG. It is a block diagram. FIG. 3 is a stage diagram showing start-up, operation and end stages of operation of the tightly coupled GPS and dead reckoning navigation system as in FIGS. 1, 2 (a) and (b). FIG. 3 is a flow chart showing the transition between various calibration states of a tightly coupled GPS and dead reckoning navigation system as in FIGS. 1, 2 (a) and (b). FIG. 6 is a flow chart of a fix mode state machine that controls the interrelationship in dead reckoning single mode, compound mode, and GPS single mode. FIG. 2 is a functional block diagram of a tightly coupled GPS and dead reckoning system for a vehicle as in FIG. It is a graph which shows the speed effect in a wheel radius. It is a graph which shows the difference of two continuous GPS course observations used by delta course calibration.

  FIG. 1 represents a GPS and dead reckoning (DR) combination embodiment of the present invention, referred to herein by the general reference numeral 100. The combination GPS and dead reckoning 100 provides navigation information on the display 102 to the user. The road map disc and player 103 displays the current user position with respect to the local road. The GPS receiver 104 tunes and tracks the microwave satellite transmission through the antenna 106 and is calibrated “delta-heading” and “delta range” calculated by the dead reckoning (DR) computer 108. -range) ”information is tightly coupled. The term “delta” represents how the course (ie, direction) of the vehicle 110 has changed over time or how the range (ie, distance) has changed over the same time period.

  The GPS receiver 104 provides data useful for calibrating the estimates communicated by the DR computer 108. The GPS receiver 104 provides absolute position and course determination, whereas the DR computer 108 can add its calibrated delta course and delta range calculation results to arrive at the dead reckoning solution. it can. The road map disc and player 103 provides road segment information to the user display 102.

  The DR computer 108 provides a very accurate dead reckoning solution that will not drift significantly if the GPS receiver has not provided a fix for a significantly long time and has lost all satellite tracking. In addition, dead reckoning solutions quickly restart the GPS receiver 104 when signal reception is restored when a large increase in position uncertainty is greatly reduced when DR is available for comparison. It is very useful to make it. This assumes worst-case movement while DR can measure vehicle movement, while a stand-alone GPS receiver promotes its position uncertainty that affects the amount of code and frequency searches. There must be. 2 (a) and 2 (b) suggest two ways in which the GPS receiver 104 and DR computer 108 can be implemented. In FIG. 2 (a), an orbiting navigation satellite transmits a microwave signal that is received by an antenna 202 and is a navigation measurement platform (MP) 204 (eg, eRide with serial output (San Francisco, Calif.)). Demodulated by OPUS-III nanoRide GPS module). The MP 204 is a type in which navigation software is run on a host processor and includes a 32-channel radio frequency (RF) receiver 206, a baseband 208, and a SAW filtering stage. The MP204 has a function of a two-channel real-time differential GPS equipped with a geostationary satellite type satellite navigation augmentation system (SBAS). The serial interface 210 can be a USB, RS-232, or similar line-based interface. The host processor 212 executes the user application 214 and has preliminary processing capability to host client software 216 with GPS, DR, and application programming interface (API) libraries. The function API 218 enables communication of navigation solutions for display to the user, and the CANbus interface 222 accepts wheel tick information from the CANbus 224.

  In FIG. 2 (b), the GPS receiver 104 and DR computer 108 are implemented with a position, velocity, time (PVT) module 250. For example, eRide (San Francisco, Calif.) EPV3600 can be used. This is a complete GPS / AGPS PVT receiver with an eRide OPUSIII baseband processor with 44,000 valid compensators and an ARM7TDMI-S (registered) with on-chip ROM and SRAM Combined with a trademark-based CPU. It can acquire and track signals down to -161 dBm, and can therefore work indoors. The ePV3600 takes a single low RF input from the ePR3036 GPS-RF front-end chip (equivalent to RF stage 206 and baseband stage 208). The embedded microcontroller (CPU) 252 operates from an external parallel flash memory, or an on-chip ROM memory or parallel flash memory. The device GPS firmware in library 216 handles acquisition, tracking, and position, velocity, and timing data output. UART 254 and serial bi-directional interface 256 support serial interface protocols that connect to “NMEA” and other applications. CANbus wheel step 258 is received by another UART 260.

  An NMEA sentence contains a first word called a data type, which defines how the rest of the sentence is interpreted. The GGA statement (Table I) is an example of essential fix data. Other sentences may repeat some of the same information, but also provide new data. Any device that can read the data can screen the data sentence it is interested in. There is no command to control the GPS, it simply transmits all the data it has even if many are ignored. The interceptor computes its own checksum and ignores any data with an incorrect checksum. The NMEA standard defines statements for all types of equipment used in marine applications. All sentences related to GPS receivers begin with “GP”.

The so-called NMEA0183 standard uses a simple ASCII serial communication protocol and defines how data is sent in “text” from a single “sender” to multiple “interceptors”. An intermediate expander allows the sender to transmit to an unlimited number of interceptors, and using a multiplexer, multiple sensors can send to a single computer port. A third-party switch is available that can establish primary and secondary callers by automatic failure bypass if the primary fails. The above standards define the content of each sentence (message) type in the host CPU application layer, so that all interceptors can parse the message correctly.

  In FIG. 1, vehicle 110 has two front, steerable wheels 112 and 114 and two rear, non-steered wheels 116 and 118. Note that for a given turn, the turning radius of the inner wheel must be much smaller than the turning radius of the outer wheel. For that reason, the calculation is complicated and depends on the steering angle, so it is not practical to derive the course and range information from the front steering wheel.

  An antilock braking system (ABS) transducer 120, 122, 124, and 126 is attached to each wheel in FIG. These transducers, sometimes referred to as wheel speed sensors (WSS), generate electronic pulses (ie, “ticks”) as the wheel turns. Some such transducers use variable reluctance and optical detectors to measure wheel rotation. The variable reluctance WSS often uses a “tone wheel” with a notch attached to the wheel rotor to produce a digital electrical output similar to a variable audio tone. If the wheel locks, the digital pulse stops and the ABS controller 128 acts to shut off the hydraulic pressure to the brake caliper for that wheel trying to recover the traction of the four tires on the road.

ABS sensor information and many other vehicle data in digital packet format are communicated by industry standard CANbus 130. Dead reckoning computer 108 receives wheel rotation step information on CANbus 130 from nodes 132, 134, 136, and 138.
Controller Area Network (CAN) is a vehicle bus standard that allows microcontrollers and devices to communicate with each other in a host computer or thousands of individual wireless vehicles. The CAN specification was published in 1991 by Robert Bosch (Stuttgart, Germany). See www.can.bosch.com/docu/can2spec.pdf

  Data traffic is multiplexed onto a single dual-wire bus (CANbus 130) and can include engine management, body control, transmission control, active suspension, automatic protection devices, environmental control and security information exchange. CAN is now the standard in OBD-II vehicle diagnostics and is required for all vehicles and light trucks manufactured in the US since 1996. The European On-Board Diagnosis Machine (EOBD) standard is similar and is mandatory for all gasoline vehicles sold in the European Union since 2001 and for all diesel vehicles since 2004.

  A transmitter in the CAN network broadcasts a message to all other nodes in the network. Each message has a unique type identifier so that the node can identify whether the message is appropriate for them. These identifiers further include a message priority field so that the CAN network can arbitrate priority between conflicting messages. A typical CAN implementation uses a twin-wire bus and has a maximum data transfer rate of 1 megabit / second. The CANbus data format has an extended error checking capability built into each data packet. The protocol automatically handles message collisions on the bus, so that higher priority messages can be moved before lower priority messages. Multiple controllers may be located on the same bus, thereby reducing the amount of wiring and the number of connectors in the vehicle 110. This also means that additional modules can be more easily added to conventional vehicle networks, such as dead reckoning computer 108.

  Any unit can start sending messages whenever the CANbus 130 is idle. If multiple units start sending messages at the same time, the bus access conflict is resolved by bitwise arbitration using the identifier. The arbitration mechanism ensures that no information or time is lost. If a data frame and a remote frame with the same identifier are started simultaneously, the data frame wins the remote frame. During arbitration, all transmitters compare the level of the transmitted bit with the level monitored on the bus. If these levels are equal, the unit can continue to send. If the “inferior” level is sent and the “dominant” level is monitored, the unit has lost mediation and must be canceled without sending any more. CANbus 130 provides a single channel to carry bits from which data resynchronization information is derived. The implementation of the channel is not indicated herein, thus allowing a single wire to be grounded, two different wires, optical fibers, etc.

  Information from the left and right wheel tick sensors 124 and 126 that are not steered is used to derive the basic delta range and delta course of the vehicle 110. Each wheel rotation produces a fixed value of the transducer pulse that is monitored by an antilock brake system (ABS) 128.

  Arithmetic average of left and right unsteered wheel rotation ticks collected by dead reckoning computer 108 in a given cycle gives accurate delta range measurements when wheel circumference or diameter is known with some accuracy be able to. The distance information obtained from the GPS receiver 104 over that same period is used for ongoing and continuous self-calibration.

  The difference in the number of rotation steps between the left and right unsteered wheels is likewise proportional to the delta course measurement when the wheel circumference or diameter is known accurately. Any error or change in the circumference or diameter of the wheels 116 and 118 will, of course, cause a corresponding error in the delta course and delta range estimates. Wheels, especially those tires, can change their diameter significantly with speed, tire pressure, temperature, wear, load, certain failures, and after-service operations.

  While a substantial step error in delta heading and delta range estimates may occur during the movement of vehicle 110, dead reckoning computer 108 or CANbus 130 is rolling without vehicle 110 moving the engine. Stopped in case. Such type of error can occur when the entire vehicle 110 is transported by trailer or ferry.

  Prior systems did not manage or control errors in the delta heading and delta range estimates very well. Embodiments of the present invention are distinguished in these respects from the prior art by the techniques described below.

  Dead reckoning computer 108 automatically integrates the delta range and delta heading measurements. It receives from the start position and course to the dead reckoning estimate expressed in packets of local level (north, east) coordinates. The initial delta range and delta heading conditions are obtained from position and velocity data routinely provided by the GPS receiver 104.

  FIG. 3 shows a combined GPS and DR system 300 that can be implemented as in FIGS. 2 (a) and 2 (b) for the application shown in FIG. The system 300 includes a GPS receiver 302 that produces a GPS positioner 304 that depends on outage and elapsed time when the antenna 306 is unable to receive microwave transmissions from GPS satellites 308 in multiple orbits. It has. Dead reckoning propagation processor 310 receives wheel ticks 312 from the wheels of the vehicle that are proportional to the number of left and right side rotations, respectively. The mode selector 314 selects whether the composite unit 316 outputs GPS alone, DR alone, or a combination of GPS and DR navigation solutions. The road map disc and player 318 provides road graphics to the user location display 320. Conventional practice is to visually “snap” the displayed user location to the nearest point on the road, even if the GPS location solution is not there.

  DR calibration 328 includes an initial ongoing calibration of wheel radius estimates and individual wheel radius differences that occur over rotation, speed, and linear operation. Dead reckoning propagation processor 310 uses these calibrations to compute the range and orientation from wheel ticks 312 that are reflected and calibrated from spare GPS positioner 304.

  When a long time elapses in the GPS positioner 304, the GPS navigation receiver 302 must normally perform a wide search to acquire the GPS satellite 308 again. Here, however, the DR range and course propagation processor 310 spends much of the GPS pause period on continuing the calculation of the position solution. Accordingly, these estimated time and position solutions are included in the instruction 330 to the GPS navigation receiver 302 to initiate the initial setup of the current DR solution search. The advantage of the instruction 330 is that it can generally be used to speed up the recapture, and can perform a more sensitive search that is more practical in high speed moving vehicle applications.

  Accordingly, the GPS receiver 302 is configured to search for parallel and independent assumptions using a number of available valid correctors (over 44,000). Some of the valid compensators in the GPS receiver 302 search for dead reckoning solutions in a very tightly coupled manner so that if the dead reckoning data is correct, the reacquisition time is minimized. If dead reckoning data is of low reliability, it can search for alternative assumptions simultaneously. In the latter case, the GPS receiver 302 can quickly enter into a dead reckoning mode 314 after a long pause or after a long power off period, for example, to correct a long unmanaged route on a ferry boat. Used to help recover.

For one-time fix (1 Hz) dead reckoning, 10 Hz wheel rotation step observations are synchronized to the same 1 second fix interval as the GPS fix produced by the GPS receiver 104 or 302. Such synchronization maintains a fixed period between fixes when the selector 314 switches between dead reckoning alone, GPS alone, and combined mode operations. The delta range and delta course based on wheel rotation increments are measured every 1 second GPS epoch. The absolute course is obtained by integrating the delta course with the initial course reference.
dead-reckoning-heading = heading-offset + sum (delta-heading)

Thereafter, any change in the east and north directions can be determined by the following equation:
Delta-East = delta-range * sin (dead-reckoning heading);
Delta-North = delta-range * cos (dead-reckoning heading)

Finally,
East position = initial east position + Sum (Delta-East);
North position = initial north position + Sum (delta-North)
It is. Dead reckoning measurement and calibration parameters are as follows.
delta-range = B _L * C _L + B _R * C _R
delta-heading = A _L * C _L -A _R * C _R
Heading = Heading-offset + sum delta-heading

here,
C _L , C _R : The number of ticks accumulated in the CANbus packet for each of the left and right wheels over a specified period such as 1000 milliseconds. The wheel rotation increment is generated by distinguishing the current and previous CANbus packets because the CANbus packet content is an accumulated number.
B _L , B _R : Calibration parameters for delta-range / tick of left and right wheels (unit: mm / tick).
A _L , A _R : Calibration parameters for delta-heading / tick difference between left and right wheels (unit: minutes / tick = angle * 60 / tick).
Heading-offset: A calibration parameter that translates sum delta-heading to a known course reference.

Thus, valid calibrations are B _L , B _R (simply remapped as “B” and dB), A _L , A _R (ie, simply “A” and dA), and heading offset (heading -offset) estimate is required. GPS receiver 104 or 302 is an accurate measure of delta-range, delta-heading, and absolute course, even when reception of that satellite transmission is not blocked Since a value can be provided, it is used as a reference source for calibration data.

The calibration parameters are initialized from a known relationship of physical attributes (eg, physical parameters) of the vehicle 110. These physical parameters, here, the left and right wheel radius _{(R l, R r: Wheel} Radius), the track width distance between the wheels (TW: Track Width), and counts or increments occurring per revolution (CPR ).

For the rotation of the vehicle 110, the distance along the turning trajectory is the product of the turning radius and the angular change (eg, changed course in radians). The turn radius is related to the road size and is unknown.
delta-range = turn radius * delta-Heading;
(Unit: meters and radians)

The generated wheel rotation increment is proportional to the product of the distance traveled and the number of steps per meter.
Count = delta-range * CPR (ticks / rotation) / [2PI wheel radius];
(Unit: tick and meter)

Therefore,
(Count / CPR) * 2PI * wheel radius = turn radius * delta-Heading
It is.

In turning to the left, for example, the turning radius of the right wheel is equal to [turning radius of the left wheel + track width]. Thus, taking the difference between the left and right delta range formulas erases the turning radius and leaves the next one.
C _L / CPR * 2 * PI * Rl- C _R / CPR * 2 * R _r = TW * delta-Heading

When rearranged, the course change in angle (delta-heading) is as follows.
delta-heading = R _l / CPR / TW * 360 * C _L -R _r / CPR / TW * 360 * C _R

The inspection is as follows.
A _L = R _l / CPR / TW * 360 (degree / tick);
A _R = R _r / CPR / TW * 360 (degree / tick)

Numerical accuracy may be improved by estimating the parameter in minutes (eg, degree * 60). Therefore, A _L and A _R are calculated as follows.
A _L = R _l / CPR / TW * 360 * 60 (minutes / tick);
A _R = R _r / CPR / TW * 360 * 60 (minutes / tick)

Here, the rotation increments of the left and right wheels represent the course change, and the average rotation increment of the left and right wheels represents the delta range at the center of the vehicle 110.
delta-range = R1 / CPR * PI * C _L + R _r / CPR * PI * C _R

When inspected
B _L = R _l / CPR * PI (meters / tick);
B _R = R _r / CPR * PI (meters / tick)
It is. Estimating this factor in millimeters improves the numerical accuracy. Therefore, B _L and B _R are calculated as follows.
B _L = R _l / CPR * PI * 1000 (millimeters / tick);
B _R = R _r / CPR * PI * 1000 (millimeters / tick)

Table II shows the physical parameters of a typical vehicle 110 with 17 inch wheel radius, 84 inch track width, and 200 steps per wheel revolution and their corresponding calibration parameters. . The example in Table II represents a realistic situation where one wheel (left) is slightly larger than the other wheel (right). The calibration parameter is a unit shown below.

  After calibration and physical parameters are obtained, an additional parameter (referred to herein as “heading-offset”) is actually needed to initiate dead reckoning. The course offset parameter can simply be determined in real time when sufficient movement of the GPS antenna 106 generates data that can be used for calibration.

The GPS course is calculated from the GPS velocity vector as atan (east / north speed) per second, true north is 0 ° and true south is 180 °. The delta course from dead reckoning is measured from the difference in wheel rotation increments measured by the parameters A _L and A _R every second. Total dead-reckoning-delta-heading-sum is the sum of all dead-reckoning delta-headings. The profile of this parameter agrees with the change in the true course, but deviates from the true course by the integration constant. Dead reckoning course is the sum of the total dead reckoning delta course and the final parameter called course offset.

  The course offset is any difference between the GPS course and the total dead reckoning delta course. The course offset can be expected to be a fixed value, but in practice it drifts depending on the stability and accuracy / availability of the CANbus data and the calibration accuracy. Any lost CANbus packets during or between fix sessions can result in unpredictable errors in the course offset. The uniqueness of each manufacturer's wheel tick sensor can also affect the stability of the course offset.

  Whenever the GPS speed exceeds 0, the GPS receiver 104 or 302 continues to update the course offset estimate. In this way, the system 100 can quickly recover from any abnormal wheel tick conditions.

  Both the heading offset and the sum of all dead reckoning delta courses must be kept as a pair in non-volatile memory, so that dead reckoning is available immediately after power on. For example, in “parking mode”, data pairs are fetched at the start of a GPS dead reckoning session. See FIG. This suggests that the data pair must be properly saved at the end of every session using a permanent stop instruction.

  “Parking start” is a mode in which dead reckoning is started immediately after the start of the GPS receiver 104 or 302, the previous calibration has been calculated, and this time, for example, is retrieved from a non-volatile memory. The parking lot start is useful when the vehicle 110 starts in the parking lot and GPS reception is not good.

  Parking start requires that a special GPS OFF command had to be executed at the previous shutdown. This OFF command is referred to herein as a “permanent shutdown” command. If received, the bits are stored in non-volatile memory with various calibration parameters indicating that a parking start is to be performed at the next GPS-on event. It is reasonable to assume that the GPS receiver 104 or 302 has not been moved between sessions. If such a prerequisite is met, the new session can be considered accurate from the beginning. This is because the previous position and course provided a valid starting condition for the new session.

  The accuracy and accuracy of the heading offset and the sum of all dead reckoning delta heading data depend on how it belongs to all CANbus data generated when the vehicle 110 is moving. Thus, after the vehicle 110 is parked and cannot move, the GPS receiver 104 or 302 should simply be turned off. The GPS receiver 104 or 302 should be enabled to run and run before the vehicle 110 moves again, so GPS can receive and buffer important CANbus data on the CANbus 130 Don't miss out.

  CANbus data must be time stamped in real time so that short pauses or gaps can be recognized and corrected. Data should not be collected and pushed into the GPS receiver 104 or 302 asynchronously without properly stamping the data. Close sharing of the GPS system and CANbus time domain is not required in embodiments of the present invention. Time stamps on the collected CANbus data are interpolated and averaged to reduce time stamp noise and to reduce the effects of any time drift between the GPS and CANbus clocks.

  Some dead reckoning embodiments can only accept 10 Hz wheel rotation increments, so the time stamp error should be less than 50 milliseconds. Recommended changes in timestamps should have a standard deviation of less than 10 milliseconds.

  The rapid availability of the GPS course is a critical factor in how quickly and accurately the system 100 can calibrate and recover from a CAN pause, wheel slip, or other abnormal event. Making the course first available is necessary to compute all other calibration parameters.

  The basic worst case course error model (in degrees) is calculated as arctan (speed error / speed). The course cannot be determined by the GPS receiver 104 or 302 when the vehicle 110 is not moving. This is because the ratio makes a round and returns. However, it is possible to determine the course at a low speed depending on the accuracy of the speed estimation value and depending on the magnitude of an arbitrary speed error. The speed error is modeled as the GPS carrier Doppler measurement noise multiplied by the horizontal accuracy loss rate (HDOP).

  Doppler measurement noise depends primarily on signal strength or signal-to-noise ratio (SNR). Any interference caused by the reflected signal in the “urban canyon” increases the true measurement error, resulting in speed and heading errors. The GPS receiver 104 or 302 is frequently operating to remove such errors from its measurements.

The conservative course error model for four different signal conditions is
Desirable model of 44dB-Hz,
34dB-Hz undesirable model,
28dB-Hz metallized cabin and range model, and
The estimated speed errors assumed to be HDOP = 6, which is another out-of-specification model of 24 dB-Hz, are 0.02, 0.2, 1.2, and 2 km / hr, respectively.

A brief overview is
Course error <44dBhz, 1 ° at 2km / hr;
Course error <34dBhz, 1 ° at 12km / hr;
Course error <28dBhz, 3 ° at 23km / hr;
Course error <24 dBhz, 3 ° at 40 km / hr.

  Embodiments of the algorithm of the present invention use these models and observe the stability of the course itself to determine which courses can be included in the calibration parameters.

  The course offset should be determined within 5 seconds as long as the speed is at or above 20 km / hr. For a metallized cabin characterized by an average SNR of 28 dB-Hz, the parameter still calibrates at a lower speed, but with reduced accuracy. However, continuous filtering can be used to reduce heading errors.

  A stop instruction, called “permanent stop”, is defined herein and referred to in FIG. It stops the GPS receiver 104 or 302 and is used as a flag to save the course calibration from one fix session to the next fix session, filling the gap over the power off period. Permanent shutdown is necessary to avoid chaos shutdown. In a conventional implementation, a normal stop command is issued and the vehicle 110 can move again before obtaining the next start command. As a result, the last position and course are no longer valid. Therefore, here the course offset is declared invalid for a general stop command. A permanent shutdown should be activated to indicate that the vehicle 110 has stopped and will not move again until the next “start” command is received.

  Some conventional code libraries need to be unable to send CANbus data to the GPS receiver 104 or 302 before initializing a GPS core with previous execution information. Any CANbus data is received before reading the calibration from the memory declaring such CANbus data that is not calibrated. CANbus data is ignored until the previous execution data is read. Here, there is no restriction on the timing of CANbus data and the reception of the previous executable file. That is, the data is processed as soon as calibration is received, and the course offset is maintained if it was previously available.

  FIG. 4 represents the CANbus data flow and timing relationship for starting and permanently stopping the GPS receiver. There are three stages 401-403 in operation. In stage 401, GPS receiver 104 or 302 is OFF, but CANbus data from CANbus 130 is present. The data can be presented to the GPS receiver 104 or 302 in the usual way, even if it has not yet started. Step 410 receives an engine ON command, which means that navigation is required. Step 412 starts the host CPU (see FIG. 7) and starts receiving CANbus data. Step 414 provides a GPS application that runs on the host CPU with a pointer in non-volatile (NV) memory to previous execution information. Step 416 loads the initial factory manufacturing parameters that are typical for the vehicle model and configuration. Step 418 starts the GPS application and hardware with a START instruction and loads it into non-volatile memory. Step 420 represents reading and writing of non-volatile memory.

  Previous execution data is not saved during dead reckoning in step 401 of the CAN flowchart. Although sector erasure is ultimately required when available storage space is consumed, such erasure makes it difficult for the dead reckoning system to keep up with the input of CANbus data. This is because erasure of the sector monopolizes the system instruction bus while erasure is being performed. Remembering previous executions during a session is meaningless. This is because the only valid storage time is when the vehicle 110 is stopped and does not move again until the next fix session. Otherwise, it is impossible to maintain the accuracy of the next parking session.

  In step 402, the GPS receiver 104 or 302 is activated and dead reckoning is available at step 422. Stage 402 ends when step 424 receives an engine OFF command.

  In step 403, the GPS receiver 104 or 302 is stopped with a permanent shutdown instruction in step 426. Step 428 represents updating the non-volatile memory with DR propagation data. CANbus data may still reach the GPS receiver 104 or 302, but they are ignored.

  FIG. 5 represents an ongoing calibration process 500 that continuously mixes various sources of calibration data with real-time calibration provided by the GPS receiver 104 or 302. This is a tight coupling between GPS and dead reckoning. The real-time calibration contribution increases the overall confidence level of the calibration status that has Status-1, Status-3, Status-4, and Status-7 levels. For example, the calibration is performed by adding 2 for parameter B, adding 1 for parameter A, and adding 4 for course offset. The highest confidence level (Status-7) is the desired state, and only dead reckoning results are available to the user.

  Don't get lost between factory parameters and partial calibration. Factory calibration inputs the previous execution data when the previous execution data is empty or cleared. Therefore, since the actual calibration starts to catch up with the expected estimate based on the initial production value, the Calibration Status-7 is not lost. Since manufacturing data is mixed with real-time data, partial calibration prior to the first full calibration is not lost. This process continues until all manufacturing estimates are overwritten with real-time data.

  In step 502, if there is no previous run or factory parameter, the dead reckoning status parameter is advanced from “0” to an initial state where the parameter is not calibrated. In step 504, the status is increased to “2”, where the calibration parameters for delta range “B” are calibrated to determine the left and right wheel radii and ratio. In step 506, the dead reckoning status parameter is increased to “3” when the calibration parameters of the delta heading and delta range “A” and the physical parameter track width (TW) are calibrated. The delta range must be calibrated before calibrating the delta course.

  In step 508, the dead reckoning status parameter is increased to “7” when the course offset, delta course and delta range are calibrated. Dead reckoning and combined mode are available at step 510 as wheel radii, ratio and track width are enabled from GPS data.

  At step 512, dead reckoning and calibration continues from its starting point if calibration is available from the previous session and is retrieved from non-volatile memory. In step 514, dead reckoning is immediately available in step 510 if Status-7 was previously reached and the previous session was terminated with a permanent stop command. If the calibration parameters are enabled by GPS data at step 518, the combined mode is available at step 516.

  Alternative calibration methods depend on the host's application programming interface (API) (see FIG. 7, eg, factory parameter mode or manufacturing mode). Most users prefer to make dead reckoning available as quickly as possible. In step 520, if the previous run data is empty or absent, such factory parameter mode provides approximate initial physical parameters for wheel radius, track width and CPR. These physical parameters are converted into A and B calibration parameters, thereby producing calibration Status-3. Only a brief observation of the GPS course at 3 meters or more per second is required to raise the calibration to Status-7, and dead reckoning navigation is then available.

  It is unrealistic to expect the highest dead reckoning accuracy (Status-7) in API factory parameter mode. This is because wheel ratio and track width cannot be fully calibrated at the factory for individual vehicles 110. The wheel radius ratio and the track width must be known to be better than 1% error. The empirical measurement provided by the GPS receiver 104 or 302 is a practical answer for obtaining a sufficiently good measurement. For this reason, the combined mode operation must be postponed until the full set of calibration parameters is accurately determined using the GPS calibration data.

  A typical sequence starting in the production mode of step 520 starts with Status-3 as the initial state. Assume that the radius and track width of the left and right wheels are calibrated. The manufacturing mode cannot specify a different radius for each wheel. Therefore, the composite mode is not available until calibration is confirmed. Status-7 occurs when the course offset is calibrated. Dead reckoning is available, and combined mode is only available after calibration is confirmed in step 518. Confirmation requires the same conditions as calibration at cold start.

There are a few minimum vehicle drive requirements necessary to operate the GPS receiver 104 or 302 sufficient to provide calibration information for parameter A, parameter B, and course offset. These are summarized in Table III. These are also reflected in steps 522 and 524 in FIG.

  Continuous calibration maximizes the life of the vehicle 110. The raw data is pre-filtered and verified before being processed by the main Kalman filter estimator for each calibration parameter. In general, good calibration can be maintained forever as long as previous run data is maintained (eg, in non-volatile memory). If the previous run data is completely lost, the original production value can be used as a starting point, so that dead reckoning can be made available as soon as possible.

There are certain conditions that reduce the calibration status to Status-7 or lower. First, as summarized in Table IV, there may be errors in the wheel rotation increments (CANbus data).

  These types of CAN errors basically degrade the calibration status to Status-3. Thus, A and B calibrations are not lost. Dead reckoning and compound modes are available immediately after errors are corrected and a new course calibration is achieved.

In addition, the calibration status may be lowered due to a calibration error. The algorithm should be constructed to minimize these occurrences until they are very rare. Advanced pre-filter logic is implemented so that the above conditions do not occur randomly or in noise calibration measurements. The most likely event that causes a calibration error is an incorrect calibration at the factory (see Table V).

  FIG. 6 illustrates the interrelationship between the fix mode state machine 600 and dead reckoning single mode 601, composite mode 602, and GPS single mode 603 depending on the availability of GPS asset and wheel data calibration. . The fix mode state machine 600 can be used to implement the mode selector 314 in FIG.

  In dead reckoning single mode 601, the test 604 checks whether a GPS fix is available from the GPS receiver 104 or 302. If yes, the test 606 asks if there is an excessively long pause (outage) of the GPS or if there is a large position change (delta position). If yes, operation switches to GPS single mode 603.

  If not, however, the test 608 asks whether dead reckoning parameter calibration is available and confirmed. If yes, combined mode 602 is allowed. Otherwise, the operation must switch to GPS only mode 603. While in combined mode 602, two tests are constantly performed. Inspection 610 checks whether two or fewer satellite vehicles are available. If true, the quality of the GPS solution is not sufficient and operation must be switched to dead reckoning single mode 601. This may occur when the vehicle 110 enters the tunnel.

  Another test that is constantly performed in combined mode 602 is a test 612 that checks whether there are many consistent satellite vehicles. If so, the GPS results are very strong and it is better to switch to GPS single mode 603. While in the GPS only mode 603, the test 614 checks to see if there are many consistent satellite vehicles. If so, the GPS result is very strong and the operation can continue in GPS single mode 603. Test 616 asks if two or fewer satellite vehicles are available and if calibration is available. If so, operation switches to dead reckoning single mode 601.

  The fix mode state machine 600 starts at step 620 when the hardware for the tightly coupled GPS and dead reckoning system 100 is first started. Test 622 determines whether a permanent stop bit has been set (eg, in non-volatile memory). If so, step 624 checks whether calibration data is available. If yes, the operation begins in dead reckoning single mode 601 and GPS receiver 104 or 302 can come out in the background without delaying the display of the vehicle navigation solution to the user, but If not, step 626 asks if a GPS fix is available. Otherwise, wait for loop 628. When GPS fix becomes available, operation switches to GPS single mode 603.

  In outdoor conditions, if the previous data including ephemeris, location and time is available and still fresh, a 1 second warm-up initial positioning time (TTFF) is possible with GPS . However, in vehicles 110 for automotive applications, the GPS signal may be shaded by vehicles in the parking lot or cannot be initiated from a GPS warm start condition. Thus, activation must begin at parking mode step 630.

  Even if the vehicle 110 is outdoors, if the ephemeris data is considered too old, a "parking lot" activation will occur, but as long as good dead reckoning data is available, for example, complete calibration 7 and the permanent stop bit are available and dead reckoning single mode 601 can begin within 1 second. Parking mode 630 is best included in systems that do not have a battery supported by a 32 kHz real time clock (RTC) that helps track GPS time better than 1 millisecond.

  If sufficient calibration or permanent stop bits are not available, parking mode 630 is not available during the session and the first fix is generated only after a new GPS fix is computed.

  In parking mode, dead reckoning always starts from the previous position. However, the accuracy of that determination cannot be determined until the first GPS fix is available.

  As soon as the first GPS fix is available, the vehicle 110 may have been moved while the GPS receiver 104 or 302 was off, so a decision about whether to trust the current dead reckoning position Must be made. For example, when the vehicle 110 is transported on a ferry boat, the delta position from the new GPS fix and the latest dead reckoning navigation are performed. For large distances, a GPS-only fix is chosen to reach an accurate position solution as quickly as possible.

  If the distance that the vehicle 110 has been moved is not large, the fix will be either a GPS only mode 603 or a combined mode 602 depending on the relative accuracy of the GPS measurements and the dead reckoning fix being communicated. A highly reliable GPS measurement or lower accuracy dead reckoning forces the state machine 600 into the GPS alone mode 603, while a less reliable GPS data imposes a more complex mode 602.

  In order to use composite mode 602, calibration must already be enabled. This is always the case when factory parameters are used. This means that the combined operation mode 602 will be short and the operation years will be short (for example, before driving sufficiently to observe the calibration capability).

  In the general case after calibration is enabled, the fix mode is dead reckoning single mode 601 when no more than two navigation satellites are available. When the number of navigation satellites is larger and the estimated multipath is lower than the estimated dead reckoning calibration accuracy, the fix becomes the GPS single mode 603. If the number of navigation satellites is not large or the estimated multipath is higher than the estimated dead reckoning calibration accuracy, the fix becomes a composite mode 602.

  The determination of the fix modes 601 to 603 is made continuously, and the selected modes 601 to 603 can be changed every second. GPS and dead reckoning conditions are always estimated to determine whether a transition to combined mode or GPS alone mode provides better results. For example, if the last mode is GPS-only mode 603, the increased multipath environment or reduction in navigation satellites is a good reason to move to combined mode 602.

  Similarly, if the current mode is combined mode 602, the condition is always estimated to determine whether the transition to GPS only mode 603 provides better results. If the current mode is dead reckoning single mode 601 and more than two satellites are available, a transition to either GPS single mode 603 or combined mode 602 occurs. For short pauses where the integral value of dead reckoning error is small, the transition is to composite mode 602. However, for longer rest periods, the GPS only mode 603 is likely to be temporarily selected to quickly snap to an unbiased position estimate. However, only a few fixes in the GPS only mode 603 are necessary to restore the best accuracy. In general, the fix mode returns to the composite mode 602 unless the number of navigation satellites is large and the estimated multipath error is not small compared to the estimated calibration accuracy.

  The information update rate is as fast as 1 Hz, providing a “super coupling” characteristic used to describe GPS and inertial navigation integrated systems. Dead reckoning and GPS information are combined in a main position / velocity Kalman filter. Dead reckoning measurements of delta range and delta heading are directly combined with GPS Doppler and pseudorange measurements.

  It provides the characteristics of the “tight” coupled approach of GPS and inertial navigation combined system 100, and GPS and dead reckoning position and velocity estimates are calculated independently and then share information to help each other It works better than the “loose” coupled solutions that are compounded together in the position / velocity region without.

  Continuous calibration continues when in combined mode with parallel calibration methodology. Independent speed and course estimators are used for continuous calibration reference of CAN calibration parameters, regardless of whether the position is calculated in GPS single mode or combined mode.

Three position modes 601-603 are used instead of only two. The basic DR input data has one-step quantization (for example, 5.6 cm for a 17-inch wheel with 48 steps). Some sensors skip outputting ticks at low speed. Pause periods, wheel slips, and missed CANbus packets may occur. B _L under check and filtering can be applied to correct most errors prior to use, but cannot be used for calibration.

  The synchronization of the wheel rotation ticks can vary depending on the CANbus 130 and host CPU load. Wheel increments are biased against banked roads. Tires can vary in size with respect to temperature, speed, and age. Dead reckoning detection and processing varies from manufacturer to manufacturer and thus has different error characteristics. Furthermore, the accuracy of dead reckoning strongly depends on the quality of the calibration. Even a 1 ° course calibration error may be integrated into a large position error for a given sufficient distance of movement. Position based on dead reckoning is essentially an integral of the derivative (delta heading and delta range). Thus, the integration will have an unknown integration constant or error from the true delta position. Thus, dead reckoning wheel rotation step integration is always biased and dead reckoning navigation cannot provide absolute positioning capability.

  GPS sensors have a completely different set of advantages and disadvantages. GPS carrier phase observation can measure better than 5% of the carrier wavelength (0.05 * 0.19m <1cm). They average the Doppler of 50 phase measurements per second from a 50 Hz phase locked loop to reduce the carrier phase standard deviation to 1 mm / sec. The integration of the GPS carrier frequency is much more accurate than trying to integrate the wheel rotation increments. The pseudo-random code phase of GPS provides an absolute position. The code phase wavelength is 293 m, but can be measured with an accuracy of 1 m in the absence of multipath interference. The vulnerability in urban areas is that GPS suffers from multipath and pure reflection.

  Given the advantages and disadvantages of dead reckoning and GPS sensors, the GPS only mode 603 is used to recover the fastest from GPS signal reception pauses that impose long periods of operation in dead reckoning only mode 601. Dead reckoning mode 601 cannot measure an absolute error and can measure only a position change. Thus, any combination of dead reckoning data simply slows recovery from GPS pauses. GPS-only mode 603 is used to determine the best absolute accuracy when the GPS situation is favorable without compounding the integration bias from dead reckoning and CANbus data errors such as wheel slip and quantization errors Is done.

  The GPS only mode 603 is used to provide a more robust solution across multiple dead reckoning systems. Each manufacturer's implementation of the CAN bus is slightly different and varies for different types of vehicles sold worldwide. Since the composite mode 602 is used only when the GPS error is larger than the dead reckoning error, maximizing the use of the GPS single mode 603 hides the difference.

  Composite mode 602 is used to stabilize GPS-based position solutions in the presence of high multipath or high horizontal accuracy reduction (HDOP) due to few usable navigation satellites. The composite mode 602 is used to maintain high GPS accuracy when entering a dense urban canyon.

  The relative weight of the composite mode 602 with respect to the GPS single mode 603 balances between the absolute accuracy of the GPS solution and the dead reckoning accuracy. This advantage can change in as little as a second.

  In general, in the field conditions, the best navigation performance is due to the GPS single mode 603, and the GPS carrier phase measurement is more accurate than the wheel sensor. In urban situations, the best navigation performance is due to the combined mode 602 because the effects of multipath and reflected noise exceed those of dead reckoning noise.

  GPS chipsets and modules without in-vehicle processors are classified as measurement platforms (MP) and provide data to the host processor through a serial UART connection. The host CPU executes navigation software that calculates position and course information. Such navigation software hosts independent of compatibility with a wide variety of host platforms including other commonly available microprocessors such as ARM, Strong ARM, Pentium (R), SH Mobile, Samsung etc. An operating system. Modules and chipsets with onboard processors are categorized as position, velocity, and time (PVT) solutions and provide sufficient navigation capabilities in standard NMEA or proprietary formats. The navigation software is embedded in a module or chipset.

  FIG. 7 represents an embodiment of the ultra-tightly coupled dead reckoning subsystem of the present invention, referenced here by the general reference numeral 700. FIG. 7 further shows details of some possible implementations of the items shown in FIGS. 1, 2 (a), 2 (b) and 3. 2 (a) and 2 (b) distinguish between implementations of functional element parts in hardware or software, where the functional interrelationship is represented in FIG.

  An orbiting navigation satellite 702 transmits a microwave signal that is received by the Navigation Measurement Platform (MP) 704 (eg, the eRide (San Francisco, Calif., OPUS-III nanoRide GPS module with serial output)). And demodulated. The MP704 has a 32-channel radio frequency (RF) receiver, a baseband, and a SAW filtering stage. The MP704 can be a two-channel real-time differential GPS with a geostationary satellite-type satellite navigation augmentation system (SBAS). These connect through hardware interface 706 and provide time report 708, fix trigger 710, and Doppler and pseudorange report 712.

  In the PVT type implementation illustrated in FIG. 2 (b), CANbus 130 data is brought in by CANbus driver 714 and UART 716, time stamped in device 718 and converted in device 720. An application programming interface (API) 722 inputs CANbus data. Instead, in the MP type illustrated in FIG. 2 (a), CANbus information with the appropriate time stamp is provided by the host application 724.

The CAN buffer 726 processes new wheel step data and buffers it for several seconds. Here, C _L and C _R are the number of ticks accumulated in the CANbus packet for each of the left and right wheels over 100 milliseconds. The number of wheel ticks is generated by taking the difference between the current CANbus packet and the previous CANbus packet because the content of the CANbus packet is the accumulated number. This data is used by dead reckoning (DR) propagation processor 730 to produce estimates of delta-range, delta-heading, and delta-heading-sum. The The DR propagation processor 730 forms a DR error model 732.

DR calibration apparatus 734, A _L, includes A _R, B _L, and a Kalman filter for B _R. The A _L and A _R calibration parameters represent the delta-heading / tick difference for the left and right wheels (eg, minutes / tick = angle * 60 / tick units). B _L and B _R are calibration parameters of delta-range / tick for the left and right wheels (for example, units of mm / tick). The course offset is a calibration parameter that translates the total delta course to a known course reference.

  The Kalman filter 736 handles delta range and delta course, as well as composite mode propagation from position, velocity, parameter B, delta range and course D, and time. The mode selection process 738 is detailed in FIG. 6 as process 600 and is represented as selector 314 in FIG.

  Data processing 742 provides the position solution encoded in the NMEA navigation message. Calibration status message 744 and NMEA message are connected to output API 746. In the PVT250 implementation of FIG. 5, the NMEA navigation message and calibration status message are output by the UART 747. Otherwise, in the MP200 type implementation of FIG. 5, NMEA navigation messages and calibration status messages are output to the host computer application 724 through the software interface.

  An application time offset calculation is provided in the offset calculation process 748 to CANbus data time, and a GPS receiver internal timer is further required for proper dead reckoning mode. This requires a 1 millisecond precision timestamp that receives both CANbus packet data and a callback function that reads the CAN timer. However, offset calibration is automatic and proceeds from receipt of the first CAN data packet. This necessary calibration step does not prevent entering dead reckoning mode with the first CANbus data packet after a valid calibration.

  GPS receiver 104 or 302 and 700 communications are defined in the NMEA specification. Most computer programs that provide real-time location information understand and expect that their data will be in NMEA format. This data includes the complete position, velocity, and time (PVT) solution computed by the GPS receiver. NMEA sends a line of data called a “sentence” that is completely self-contained and unrelated to other sentences. There is a standard statement for each device category, and there is also the ability to define dedicated statements used by individual companies. All standard sentences have a two-letter prefix that defines the device that uses the sentence pattern. For GPS receivers, the prefix is GP. The prefix character is followed by three consecutive characters that define the content of the sentence. NMEA allows hardware manufacturers to define their own sentences starting with the letter P and followed by three characters that identify the manufacturer. For example, Garmin's sentence begins with PGRM, and Magellan's sentence begins with PMGN.

  Each sentence can start with a “$”, end with a carriage return sequence, and can be up to 80 characters of visible text (plus a line terminator). Data is contained within this single line with data items separated by commas. The data itself is simply ASCII text and can extend over several sentences in some specialized examples, but is usually contained entirely in one variable length sentence. Data may vary in the amount of accuracy included in the message. For example, the time may be indicated as a fraction of a second, or the position may be indicated with 3 or 4 digits after the decimal point. Programs that read data should simply use commas to determine field boundaries and are independent of column position. There is provision for a checksum at the end of each sentence, which may be checked by the unit reading the data. The checksum field consists of one “*” and two hexadecimal numbers representing an 8-bit exclusive OR between all characters, but does not include “$” and “*”. A checksum is necessary for some sentences.

  Although there have been some changes in the above standards, the only ones that may be encountered with the use of GPS are 1.5 and 2.0-2.3. These simply specify a number of different sentence constructs, which may be specific to the needs of a particular device. Thus, the GPS may need to be changed to suit the connected device. Some GPSs provide the ability to compose a customized set of sentences, while other GPSs can provide a fixed set of options. Many GPS receivers simply output a fixed set of sentences that cannot be changed by the user. The current version standard is 3.01. I don't have any specific information about this version, but I don't know any GPS products that require conformance to this version.

  NMEA consists of a sentence, and its first word, called the data type, defines the remaining interpretation of the sentence. Each data type has its own specific interpretation and is defined in the NMEA standard. The $ GPGGA statement provides essential fix data. Other sentences may repeat some of the same information but also provide new data. A device or program that reads the data can focus on the data sentences that it is interested in, but simply ignores other sentences that are not of interest. In the NMEA standard, there is no command to indicate that GPS should do something else. Instead, each receiver simply sends all the data and expects many to be ignored. Some receivers have instructions in the unit that can select a subset of all sentences to send, and in some cases individual sentences. There is no way to indicate to the unit whether the sentence has been read correctly or to request to resend some data that is not available. Instead, the receiving unit simply checks the checksum and ignores the data if it thinks the checksum is bad and the data will be transmitted again later.

  Dead reckoning fixes are reported by GPS receivers 104 or 302, and 700 in both ASCII format NMEA strings, as well as the function API. Dead reckoning fixes are indicated in NMEA format in the $ GPGGA string by a fix indicator value of 6. The GPS fix value is 1. Dead reckoning fixes are reported by values in the function API. Dead reckoning also provides a speed estimate by dividing the delta range by a known observation period. Both of these parameters are visible in the NMEA $ GPRMC string and also the function API. These fields report only dead reckoning fixes after valid calibration is achieved.

The calibration status can be seen in the string. Dead reckoning debug messages are available in standard NMEA format useful for detailed monitoring of dead reckoning calibration progress and status. For example, see Table VI. It is output from the core library with standard NMEA output, but the message must be sought and then sent to an output port or file.

  There is no end time associated with any saved calibration. A new calibration starts as soon as GPS / dead reckoning is enabled by command and CANbus packets are available. The status level of the last calibration is not affected by the calibration epoch.

Knowing the physical parameters for the vehicle 110 allows the initial calibration to enter the calibration algorithm through the API 722. The A and B parameters are initialized according to the left and right radii (R ₁ , R _r ), the number of rotations per revolution (CPR), and the track width (TW). Dead reckoning mode 601 is not possible until the first GPS fix is made at sufficient speed, so that a good course offset can be established.

  API 722 provides a manufacturing method to initialize system 700 and speed it up when calibration is available, and API 722 and output API 746 allow functional system testing of CANbus packet processing 726 and 730.

  Since the same application thread is used at the beginning of each session, data indicating which source of CANbus data is used is typically provided by the host CPU application for each GPS session. In many cases, this involves sending physical parameters. It is assumed that the physical parameters entered by the manufacturer cannot be as accurate as anything estimated during operation by the running GPS receiver 104 or 302.

  Accurate calibration requires knowing the physical parameters very accurately. Even small apparent effects of tire pressure can significantly affect the actual wheel radius and vary between day and night. For this reason, factory physical parameters should only be loaded when a previous run of GPS calibrated data has been cleared. This means that previous GPS-based calibrations are not overwritten with factory parameters through the API.

  In most manufacturing cases, it is expected that the manufacturer will not know the actual wheel radius with sufficient accuracy to provide reliable dead reckoning. For example, the tire pressure may not be full or the pressure on each wheel may not be equal.

  If the user wants to be available quickly and can provide high-precision calibration, GPS will be available, and dead reckoning can be used as soon as the vehicle 110 is moved. If the manufacturer can purely provide a reasonably accurate calibration (within a few percent), the accuracy penalty is small before a full GPS-based calibration installs itself. This is the case for airbag deployment applications, where the availability of dead reckoning is initially more important than sufficient accuracy.

  In the low-accuracy factory mode, the physical parameters are used as calibration initial conditions and classified as accurate inside, but the calibration reliability observable from the outside is set to zero. Dead reckoning is available after GPS-based calibration improves the initial parameters and completes the normal calibration process. This mode is selected to speed up calibration and not allow dead reckoning until GPS calibration is complete. The time to effective dead reckoning calibration is faster than if no starting value was provided.

  Laboratory testing mode provides functional verification of software and systems and is used for laboratory testing. Before and after the inspection is completed to avoid problems in other factory modes, the previous executable is deleted. Dead reckoning is immediately available without any additional calibration. The starting course always starts from 0 and requires a starting position. The location can be installed via the location input API after sending physical parameter instructions or by obtaining a GPS fix.

  Inspection vehicle calibration can be transferred to a family of similar vehicles, where immediate dead reckoning is a higher priority than having initial accuracy. If dead reckoning is needed immediately and it is not possible to calibrate each vehicle individually in the factory, a single vehicle is selected as the representative of the family of vehicles. It is configured for factory default mode. Initial calibration is not required and the test driver can run a calibration suite to accumulate GPS-based calibrations. PERDCR, 11 NMEA messages such as Table VI are enabled and recorded during calibration performed on the storage computer. The NMEA file is retrieved from the storage computer and the A and B calibration parameters are extracted from the final PERDCR, 11 string. The physical parameters are estimated from the calibration parameters. The physical parameters are loaded at the factory during installation and the factory high-precision calibration function is selected. The host CPU application is configured to read physical parameters from the file or they are hard coded into the host CPU application. The vehicle family can then observe rapid dead reckoning as soon as the first GPS fix is obtained at a sufficient speed (eg, 25 km / h or more).

Simple mathematics is used to estimate physical parameters from calibration parameters. The estimated calibration parameters are easily retrieved from the dedicated NMEA string PERDCR, 11. The calibration parameters observed in PERDCR, 11 have the following units and sizes.
A _L = R _l / CPR / TW * 360 * 60 (minutes / tick);
A _R = R _r / CPR / TW * 360 * 60 (minutes / tick);
B _L = R _l / CPR * PI * 1000 (millimeters / tick);
B _R = R _r / CPR * PI * 1000 (millimeters / tick)

The estimated calibration parameter from PERDCR, 11 is defined by:
A _L '= estimate of true A _L ;
A _R '= estimate of true A _R ;
B _L '= estimate of true B _L ；
B _R '= estimate of true B _R

The physical parameter (TW) of the track width is estimated by the following equation.
TWl '= B _L ' / A _L '* (360 * 60) / (PI * 1000) (meters);
TWr '= B _R ' / A _R '* (360 * 60) / (PI * 1000) (meters)

The calibration algorithm ensures that the B / A ratio for both wheels is the same. Either left or right wheel parameters can be used. Another approach is to average these two estimates, resulting in:
TW 'estimate = (TWl' + TWr ') / 2

Unfortunately, the wheel radii (R ₁ , R _r ) cannot be separated from the number of revolutions per revolution (CPR). These parameters are always combined as in the above formula. Each vehicle manufacturer supplies CPR in a set of physical parameters. In this case, the wheel radius is estimated as follows.
R _l = B _L '* CPR / PI / 1000;
R _r = B _L '* CPR / PI / 1000

Embodiments of the real-time calibration method of the present invention have three steps: (1) compare the delta range with GPS speed and estimate B _L and _BR parameters; (2) Given, compare delta course with changes in GPS course and estimate A _L and A _R parameters; and (3) Give A and B parameters and use GPS course as absolute reference to estimate course offset Is done.

  Before a dead reckoning fix can be computed, there must be (a) a valid calibration of parameters A and B, heading offset, (b) a valid position, and (3) Input of CANbus data packets with synchronized time stamps.

  The GPS receiver 104 or 302 does not require an estimate of GPS time to operate the dead reckoning single mode 601. This allows the “parking mode” to proceed even if the GPS receiver 104 or 302 is not providing time information. A valid calibration as represented by previous execution data 750 in FIG. 7 can be provided either from non-volatile memory data or input through an API.

  Dead reckoning calibration starts automatically, for example, as soon as any CANbus packet is received, without initialization from the API command. The time for available calibration depends on the availability of GPS data and the completion of straight ahead and turns including large course changes. The time to be calibrated is reduced by the higher speed. This is because more steps can be observed and the quantization effect is proportionally weakened.

  Calibration of the B parameter, wheel radius, and ratio requires either a straight line segment or straight line travel of the GPS receiver 104 or 302, where the course changes, such as figure 8 or straight line travel. Canceled. The minimum requirement is to run at least 5 such segments. Thereafter, the A calibration calculates the track width, and then proceeds to execute the turn. Right angle turning is better than at least 5 circles. The course offset can then be calibrated after parameter A and parameter B are obtained.

  Thus, the minimum requirement to collect and generate calibrations is to obtain at least 70 meters of straight segments without significant acceleration, at least 5 right angle turns, and at least 3 meters per second for 5 seconds to obtain a course offset. It is a straight line segment. Such an example has an undesirable signal level of 34 dBHz.

  Wheel rotation step sensors 120, 122, 124, and 126 (FIG. 1) are generally hall-type devices, such as a digital clock step when the wheel rotates and the magnetic bumps pass the detector. Create electrical nicks. The number of ticks that occur for one revolution of the wheel ranges from 48 in Peugeot cars to 98 in Mercedes cars.

  The original string of wheel rotation ticks produced by the sensor is continuous, but CANbus 130 must inevitably bundle the wheel ticks into 1-10 millisecond data packets for multiplexing. . Secondary processing groups these shorter packets into longer packets, each over 100 milliseconds (eg, 10 Hz rate). These 10Hz CAN wheel rotation step packets are accessible by external applications through a protective gateway between the secure life safety network and the external application network.

  CANbus 130 has a processor (CPU) with a system clock that is used to generate a sampling range of wheel rotation increments. Such CPU clocks are generally not available externally for synchronization, so the time domain between 10 Hz packet generation and any GPS receiver clock must be handled asynchronously.

  In embodiments where the CANbus and GPS receiver clocks can be synchronized, carrier Doppler, code phase, and other GPS observations can be taken over the same time range. It is a good way to synchronize from master to slave with 1 pulse / second (1-PPS) pulse. Another embodiment uses the same clock or time difference or frequency difference circuit to observe the relative time or frequency difference between the two clock domains. The OPR ™ baseband chip from eRide (San Francisco, Calif.) has such capabilities, but its use dramatically complicates system integration.

  Synchronization of CANbus data is further complicated by the delay inherent in the transmission of packet messages over networks such as CANbus 130 and GPS receiver 104 or 302. Typically, the third CPU is physically connected to CANbus 130 and converts from a CAN protocol to a dedicated protocol. This is because all manufacturers use different CAN message protocols and wheel ticks. Such a CPU converts to a common format for the CAN input of the GPS receiver. There are three types of delays: (1) generated from CANbus 130 delay, (2) generated from CAN format converter delay, and (3) receive buffer included in GPS receiver 104 or 302. Outbreak. The combination of these delays is called “CANbus jitter” and has an invariant part and a variable part.

  The combination of CANbus jitter and any time drift between the CANbus clock and the GPS clock creates a noise factor in the synchronization and processing of CAN and GPS data.

  Synchronization is obtained by stamping each CANbus packet as it arrives at the GPS receiver's buffer through a universal asynchronous transceiver (UART) serial interrupt driver.

  The GPS receiver software running as a host CPU program must select which packet is closest in time to the GPS time range in order to use CANbus data for both calibration and dead reckoning propagation. A typical range is 1 second. There are usually 10 packets with a time tag within the 1 second time range of each GPS.

  In the Mercedes system experiment, the time tag shifted rapidly and in just a few minutes, actually 11 packets, indicating that the CANbus clock was running fast, or 9 packets. Indicates that the CANbus clock operates at low speed. Furthermore, CANbus jitter is seen to be responsible for 9 or 11 packet variations, typically with a 1 second acquisition operation.

  Such a problem has a simple solution. First, the offset between the start time of a noisy packet and the GPS range time is filtered to remove most jitter. The average difference is maintained. It is biased with a drift component, so the filter is configured to track changes. The original time tag with jitter is used to select packets that include both start and end intervals in the GPS time domain. The average difference is then used to define the portion of the packet that has a GPS fix time. The contribution of packets within the GPS interval is interpolated from the total number of packets.

  For example, in a packet having a GPS start time, a portion earlier than the GPS start time and the scheduled packet time is interpolated, leaving a portion generated after the GPS time by subtracting the average difference. Similarly, for packets that have data that occurred after the GPS end time, interpolation assumes the portion used and uses the average difference from the GPS end time to the scheduled packet start time. Remove any period after time. Other packets between these two boundary packets are used in their entirety.

  Such a technique is used for ticks from both the left and right wheel sensors 124 and 126, so that the number of interpolated wheel rotation ticks is adjusted to best match the corresponding GPS time interval. .

  The GPS start time and end time are different for each formula of delta range and delta course. In the delta range measurement, GPS speeds of 1 second or more are compared at the same interval with all wheel rotation increments. Since the GPS carrier tracking wavelength is 0.1904m, a one second integral of the GPS velocity is used, which is determined within a few degrees using a conventional phase-locked look method common to GPS-based velocity determination. . On the other hand, the GPS code phase has a wavelength corresponding to one chip of 293 m. Therefore, the 1 second delta range from the code phase is too noisy for a short time calibration.

  Long-term calibration with GPS derived position solutions is possible, but they require tremendous processing and memory.

  The preferred embodiment uses east and north speeds from the GPS speed solution as the main calibration source. The delta range from GPS is defined as the square root of the sum of the east speed squared and the north speed squared. The course is defined as the quadrant result of the inverse tangent (arc tangent) of the eastern speed divided by the northward speed. The east and north velocities include both positive and negative signs and magnitudes.

  The GPS east and north speeds themselves are estimated from GPS Doppler measurements. The Doppler measurement is the average of the carrier Doppler frequency divided by the measurement interval. The applicable time for the measurement is the center of the above range.

  In delta range calibration, wheel rotation increments are selected by start and end times that approximate the start and end times of GPS measurements.

  The velocity fix is made at the center of the one second range defined by the time tracking state machine (TSM) measurements. The GPS fix course applies to the center of the TSM range. Timing is assumed that the transition from GPS to DR does not cause time tag jumps.

  Left increment (TL) and right increment (TR) are defined as 1 second wheel increments. In the course, TL and TR have a time range from the center of the previous range to the center of the new range.

The coefficients are as follows.
A _L = RadiusLeft / CPR * TRACKWIDTH * degree * 60 (minutes / tick);
B _R = RadiusLeft / CPR * 1000 (millimeters / tick)

The one second delta course is as follows.
δh = A _L * TL-A _R * TR

And
A _L = A + dA
A _R = A-dA
δh = (A + dA) TL-(A-dA) TR = A * [TL-TR] + dA * [TL + TR]
When DH = 0, the arc is closed for a while or is running straight, as judged from GPS measurements.

next,
A _L * TL = A _R * TR
A _L / A _R = TR / TL
α = A _L / A _R = (A + dA) / (A-dA)
Make (1-α) / (1 + α) and solve as
(1-α) / (1 + α) = -dA / A
Therefore, dA = -A (1-α) / (1 + α)

In the delta range, the wheel increment is defined as the raw data increment over the same time frame as the TSM range.
δh = A * [TL-TR]-(1-α) / (1 + α) * [TL + TR]
= A * [(TL-TR)-(1-α) / (1 + α) * (TL + TR)]

The 1 second delta range is as follows.
dr = B _L * TR + B _R * TR

And
B _L = B + dB
B _R = B-dB

as a result,
dR = (B + dB) * TL + (B-dB) * TR
= B * (TL + TR) + dB * (TL-TR)

By the definition of B _L and B _R in the physical parameters,
B _L / B _R = A _L / A _R

Therefore,
dB = -B (1-α) / (1 + α)
dr = B * (TL + TR)-B (1-α) / (1 + α) * (TL-TR)
dr = B * [(TL + TR)-(1-α) / (1 + α) * (TL-TR)]

To calculate dead reckoning fix that propagates the position by 1 second from the previous fix time,
deltaN = dr * cos heading (delta-north);
deltaE = dr * sin heading (delta-east);
heading h (ti) = h0 + Σdhn (n = 0, i)
It is. _{A L, A R, B L} , and assuming that the estimated notarized calibration B _R,
1) To estimate α, find the orbital segment where + Σdhn (n = 0, i) GPS is 0. In this case, the ratio of steps from right to left is defined as follows.
a.α = TR / TL
2) In order to estimate B, the same wheel increment is obtained from the GPS estimated dr value and the same observation range as the GPS range change observation range.
B = dr / [(TL + TR)-(1-α) / (1 + α) * (TL-TR)]
3) To estimate A, obtain an estimate of dh from GPS, and obtain CAN increments from the same observation range as the GPS course change observation range.
A = δh / [(TL-TR)-(1-α) / (1 + α) * (TL + TR)]
4) To estimate h0, keep the total (dh) and GPS heading difference if the GPS heading is reliable.

  DR is stable for a short time and unstable for a long time. GPS is unstable for a short time and stable for a long time.

  Calibration is obtained using point observation. These observations are each derived from velocity information that is an intermediate value of 0, but the integration is not stationary. To get the final calibration at rest, information from the GPS location is needed.

For long-term observation of dE and dN from GPS,
DE _gps = ΣdE _gps
DN _gps = ΣdN _gps

With a similar estimate formed from a short-calibrated DR,
DE _dr = ΣdE _dr
DN _dr = ΣdN _dr
It is. For a similar estimate formed from a short time calibrated DR, a total of 3 points is used, for example.
DE _dr = ΣdE _dr
= dr1 * sin (h0 + dh1) + dr2 * sin (h0 + dh2 + dh2) + dr3 * sin (do + dh1 + dh2 + dh3);
DN _dr = ΣdN _dr
= dr1 * cos (h0 + dh1) + dr2 * cos (h0 + dh2 + dh2) + dr3 * cos (do + dh1 + dh2 + dh3)
The estimates of A and B are assumed to be sufficiently accurate so that a first order Taylor series can be used.

Embodiments of the present invention compensate for rotational ticks for the speed effect phenomenon represented by curve 800 in FIG. The low speed segment 801 below the low speed cutoff is an essentially flat line. The intermediate speed segment 802 between the low speed cutoff and the high speed cutoff follows a polynomial, 1 + c _1st (s-SL) + c _2nd (s-SL) ² . The high speed segment 303 exceeds the scale limit of the coefficient.

  Speed related effects result in a change in the difference in the number of left and right wheel ticks in rotation with respect to speed due to changes in wheel radius. In other words, the same rotation obtained at different speeds causes a difference in the number of ticks between the left and right wheels. The cause is wheel warping and vehicle suspension reaction.

  At higher speeds, a greater amount of turning energy, seen as wheel increment differences, occurs compared to lower speeds. Of course, the higher the speed, the more ticks per second, so the wheel must compensate for the apparent angular speed calculated from the rotational ticks.

The turn ticks that enter the individual basic delta course updates must be properly compensated, where:
delta Heading = A * turn ticks
It is. The rotational increment is routinely compensated based on the difference in wheel radius when going straight.
Delta heading = A _left * WheelTicksLeft-A _right * WheelTicksRight
= A * (WheelTicksLeft-WheelTicksRight)-alpha * (WheelTicksLeft + WheelTicksRight)
= A * compensated turn ticks
= A * compTT
It is. Rotation not compensated for
turnTicks = left wheel ticks-right wheel ticks
It is. The low speed correction (1) for segment 801 relative to the wheel radius is
compensated turn ticks (1) compTT (1) = turn ticks-alpha * sum ticks;
alpha = 1-ratio / 1 + ratio
Where the “ratio” is the filtered value of the right wheel increment divided by the left wheel increment when not turning.
The correction of the segment 802 to the speed effect CompTT (2) measures compTT (1) by the following polynomial.
1 + c _1st (s-SL) + c _2nd (s-SL) ²
This is a function of total increments, where
s = sum ticks
SL = below reference point low speed cutoff, polynomial = 1
Above the fast cutoff, polynomial = coeffScale limit.

  In delta course calibration, the difference between two consecutive GPS course observations is used, as represented in FIG. The time tag for each course is the center of the GPS measurement range. Thus, the difference between the two GPS courses has an applicable time between the two time tags of the two GPS courses. For this reason, the wheel rotation increment used for the delta course is in a different time range than that for the delta range, and the delta range is the higher speed used to form each course individually, Or it can either be 0.5 seconds ahead of the delta range range or 0.5 seconds behind, depending on whether it is at a slower speed.

  In practice, the faster (or slower) speed1 is used for the delta range. This is because there may be a delay in receiving all the CANbus packets necessary to extend the speed2 interval, so that speed2 is on the CAN communication line (eg CAN to converter to GPS serial port) Because it requires more rapid data.

In general, the basic formula for rotating the left and right wheels uses a scale factor that simplifies each wheel.
DH = delta-heading = A _left * ticks left for DH-A _right * ticks right for DH;
DR = delta-range = B _left * ticks left for DR + B _left * ticks right for DR

further,
DR heading = initial heading from GPS + sum DH;
When
Delta East = DR sin (DR heading);
Delta North = DR cos (DR heading)
It is.

  Experiments show that parameter estimation is not as simple as it looks. Parameters A and B are no longer constant. They change with time. This is because the wheel radius changes with tire pressure, external temperature, speed, turning effect, vehicle load, passenger arrangement and number of people, tire wear, and the like. The solution described here models each parameter as a slowly varying parameter, each corresponding to a different time model.

  Although it is possible to solve all the parameters in the multivariable nonlinear system mathematical formula, it was found that the parameters can be independently calculated by a single variable mathematical system if they are performed in the correct order.

Instead of solving each of the left and right parameters for A and B,
A _right / A _left = B _left / B _right = r

The 1-second delta course is
δh = A _L * TL-A _R * TR;

And
A _L = A + dA;
A _R = A-dA;
δh = (A + dA) TL-(A-dA) TR;
= A * [TL-TR] + dA * [TL + TR];

And r = A _L / A _R = (A + dA) / (A-dA);

Once sorted,
dA = -A (1-r) / (1 + r)
It is. Therefore,
delta-heading = A * [(TL-TR)-(1-r) / (1 + r) * (TL + TR)]
= A * compTT
It is. Here, the compensated rotational step energy is
compTT = (TL-TR)-(1-r) / (1 + r) * (TL + TR)
It is. Similarly, solving the common B parameter gives
delta-range = B * [(TL + TR)-(1-α) / (1 + α) * (TL-TR)]
= B * compST
It is. Where compST = compensated total increment.

  A simple formulation deals with the general case where the left and right wheels do not effectively have the same radius. The bias between the left and right wheel increments accumulates even when the vehicle 110 is traveling straight.

A simple way to estimate this parameter is to use the GPS receiver 104 or 302 to isolate driving conditions in which the vehicle 110 goes straight or makes a slight turn. Under this condition, the ratio “R” can be estimated as follows over a period in which any change in the GPS course from the start of integration to the end of integration is a small change in the GPS course.
R = sum ticks right / sum ticks left

  That is, the start and end periods are considered when the absolute value of (headGPSstart-headGPS) is typically below a threshold of 1 °.

  The estimate of the delta course includes any true change in the course of the vehicle 110 resulting in any GPS course error and accurate turning energy in the wheel rotation increment.

  To provide this and all common change states, the individual r estimates are blended with a single state Kalman filter. The noise model for each measurement reveals the reliability of the two GPS course estimates. The processing noise model for the Kalman filter is adjusted to provide how fast the ratio parameter can change.

  The processing noise is always set large when the vehicle 110 is stopped for a long time. This is because changes in vehicle load can significantly affect the tire radius. Uniformizing the processing noise to a steady state allows the ratio estimate to provide a change in condition every 5 minutes.

  When operating in urban canyons and / or very weak signal conditions, GPS course estimates can be unstable and unreliable. A very large noise model must be used. The deselection scheme is used to select out-of-range measurements from the corresponding Kalman filter that are too far from normal values.

  Such a ratio is always estimated when the vehicle 110 goes straight or the net course change cancels out to zero (eg, 15 ° right turn followed by 15 ° left turn). be able to. The ratio compensates for wheel radius differences and straightens the run. Applying a ratio of 1.0 blindly allows dead reckoning propagation to plan a route with a large curve, even when traveling on a straight road.

Experience has shown that the terms A and A _left and A _right may deviate from reality during a turn, depending on how the vehicle 110 suspension operates. In general, more turning energy is observed that gives the impression of oversteering (ie, ground tracking shows more than a true course change).

  To compensate for this effect, a speed compensation parameter “C” is introduced here to further compensate for the rotational step already compensated by any actual difference in wheel radius. The general form of the C parameter is the value 0: 1. At low speed, the C parameter is near 0, and at high speed it converges but does not reach 1.

  The C parameter is more difficult to estimate. This is because it requires observation of turning conditions over the entire speed range possible for the vehicle 110. Therefore, a baseline or default model is used. The C parameter is frequently improved with another single state Kalman filter over the life of the vehicle 110.

  A typical form of the C parameter is a unit from zero speed to a baseline value called LowSpeedCutOff. The curve then decreases and becomes linear to the maximum speed at which the curve stays flat at a fixed lower value. The slope of the intermediate region is estimated in real time. The independent variable for simplifying the speed representation is the sum of the left and right wheel rotation increments, here called sumTicks.

The modeling formula for estimating the C parameter is
deltaHeading = A * (1 + C * (compST-LowSpeedCutOff)) * compTT
It is. An estimate of the C parameter can be formed as follows.
C estimate = (deltaHeadingGPS-A * compTT) / (A * compTT * (compST-LowSpeedCutOff))
A baseline estimate of the A parameter is required, and the C parameter has no units.

The A parameter converts the change in wheel rotation increment between the left and right wheels into a course change. The course change is then integrated with an arbitrary constant to produce a total dead reckoning course. The calculation of the A parameter requires collecting at least two GPS speed estimates that can obtain a delta course, and collecting a compensated rotational step estimate corresponding to the same integration period. Therefore,
A-estimate = deltaHeadingGPS / (compTT * C)
It is. Here, the A estimate (A-estimate) is radians per rotation increment.

  Experiments show that the accuracy depends on the quality of the GPS course measurements. Software programming should have at least two ways to estimate the course. For example, the least square batch estimator estimates the east and north velocities from a set of GPS Doppler measurements obtained in one time interval, so it does not have a dynamic model and is therefore delayed under dynamics. Does not have. In another example, the estimator uses a navigational Kalman filter with a complex measurement error model for each GPS Doppler measurement and a complex model of how GPS speed and position change from epoch to epoch. I have. Under weak signal conditions with higher noise ratios, the Kalman filter provides less variability in east and north velocities.

  Another single state Kalman filter is used to blend together the time series of A estimates. Each A estimate has a measurement error model based on an error model in the GPS course component. A processing noise model is included that sets a filter time constant that slowly adapts to changing tire conditions.

  The A estimate depends on two other parameters, the wheel radius ratio r, and the speed compensation model C. The A parameter is estimated continuously from the initial actual behavior and uses one value for the ratio and default model until the parameters can be estimated respectively.

In estimating parameter B, delta range parameter B is used to set the speed for dead reckoning. It converts the sum of left and right wheel rotation increments to dead reckoning speed. Parameter B is estimated using the GPS speed and the compensated total step measurement (compST) as follows:
B-estimate = GPS speed in one second / compST in one second

  The B estimate is blended with a single state Kalman filter, where the measurement noise is a function of the GPS speed error estimate. The processing noise for the Kalman filter sets a time constant for how B varies with time. In general, tires compress and expand at various speeds and vehicle loads. Accordingly, the amount of processing noise to be included is set to have a time constant of about 5 minutes.

For the course offset O parameter, it is necessary to have an absolute course estimate from dead reckoning to estimate the change in dead reckoning position from a set of new left and right wheel rotation increments. This parameter is called DR_heading here.
DR_heading = heading offset + sum of all delta-headings

The course offset is set first when the first reliable GPS course becomes available (eg, as soon as the GPS receiver moves at a speed of several meters per second or more). To initialize
Heading offset = Heading from last fix
It is. And to update
DR heading on first update = heading offset + deltaHEading1
It is. In the second update,
DR heading = heading offset + deltaHeading1 + deltaHeading2;
= heading offset + sum of all DR deltaHeadings;
= heading offset + sumDeltaHeading
It is.

If the sum of all such delta course estimates is complete, the course offset is a fixed constant and does not need to be updated. In practice, however, the heading offset may change fairly rapidly due to system noise and modeling failure. In order to continuously estimate the course offset in order to keep the course offset accurate, another signal state Kalman filter is required.
The heading offset estimate, O = GPS heading-sumDeltaHeading
Then, it is blended with all the course offset estimated values corresponding to the measurement noise model based on the GPS course error estimated value. The time constant is set at about 5 minutes.

In summary, calibration proceeds as follows in Table VII.

  One embodiment of the present invention comprises a tightly coupled GPS and dead reckoning software application program for execution as a GPS application in a host computer. It is connected to receive measurements from a GPS measurement platform (MP) on a vehicle with wheels. For example, MP 704 (FIG. 7) on a vehicle 110 (FIG. 1) with wheels 112, 114, 116, and 118. The tightly coupled GPS and dead reckoning software application program includes input processing for collecting time, fix triggers, Doppler, and pseudorange reports from the GPS measurement platform. The offset calculation process from GPS time to host computer time is used to align otherwise asynchronous data generated by the vehicle to GPS time. Least square snapshot estimates include position (P), velocity (V), calibration parameters for wheel radius and ratio (B), delta heading and delta range (D), and individual new position fix and GPS system time Is calculated for the time based on The Kalman filter is included and configured for position (P), velocity (V), calibration parameters for wheel radius and ratio (B), delta course and delta range (D), and time, and Process and propagate dead reckoning delta range, delta heading, combined mode.

The continuous mode selection process selects operation in GPS single mode, dead reckoning single mode, or a combined mode of both GPS and dead reckoning, depending on the availability of GPS solutions and wheel data calibration. The new fix computation process computes hardware advance data and adjusts the GPS millisecond timer connected to the time offset computation process. The buffer processor collects CANbus data stamped with time stamps for wheel rotation ticks measured by the vehicle wheel transducer. Dead-reckoning calibration Kalman filter includes A _L and A _R calibration parameters for delta-heading / tick difference between left and right wheels, and B _L and B _R calibration for delta-range / tick of left and right wheels Process the parameters as well as the course offset calibration parameters to translate the total delta course to a known course reference. Dead Reckoning Data Processor converts data from the buffer processor to the delta range, delta heading, and delta heading sum information for later processing by a Kalman filter and includes a dead reckoning error model . The application programming interface (API) outputs a new navigation fix.

  Tightly coupled GPS and dead reckoning software application program operates in one of three modes including GPS receiver alone, DR computer alone, and combined mode including both GPS receiver and DR computer A selection process may further be included. Operation in the GPS receiver only mode occurs when a GPS fix is available, there are satellite vehicles that are available sufficiently consistently, and more than two satellite vehicles are being tracked. Operation in DR computer single mode occurs when GPS fix is not available or when two or fewer satellite vehicles are being tracked. And operation in combined mode confirms from inspection that no more than two satellite vehicles are being tracked, or that there are no satellite vehicles available consistently, and that calibration is available. It happens when it is done. The new navigation fix is then output through the API.

In summary, Table VIII describes the parameters for wheel ticks in CAN frames, and Table IX lists the speed compensation parameters set by the API.

The algorithm is as follows.
The model formula for estimating C1 is
dh = A * (1 + C1 * (SumT-LowSpeedCutOff)) * dT
It is. Once sorted,
C1 = (dh-A * dT) / (A * dT * (SumT-LowSpeedCutOff))
It is. here,
TL: Left wheel ticks
TR: Right wheel ticks
A: Parameters for calculating the delta course
Ratio: Ratio of left and right wheel increments estimated for the formula "Ratio = TL / TR"
SumT: Sum of wheel increments is defined as “SumT = TL + TR”
dT: Wheel tick difference is defined as “dT = (TL−TR) + (1−Ratio) / (1 + Ratio) * SumT”.

  While particular embodiments of the present invention have been described and illustrated, it is not intended to limit the invention. Modifications and changes will no doubt become apparent to those skilled in the art, and it is intended that the invention be limited only by the scope of the appended claims.

100, 300 GPS and inertial navigation combined system 102 Display 103, 318 Road map disc and player 104, 302 GPS receiver 106, 306 Antenna 108 DR computer 110 Vehicle 112, 114 Two steerable wheels 116, 118 in front Rear two unsteered wheels 120, 122, 124, 126 Anti-lock braking system (ABS) transducer 128 ABS controller 130 CANBus
132, 134, 136, 138 Node 202 Antenna 204 Navigation Measurement Platform (MP)
206 32 Channel Radio Frequency (RF) Receiver 208 Baseband 210 Serial Interface 212 Host Processor 214 User Application 216 Client Software 218 Functional API
222 CANbus interface 224 CANbus
250 Position, Speed, Time (PVT) Module 252 Microcontroller (CPU)
254, 260 UART
256 Serial Bidirectional Interface 304 GPS Positioner 308 GPS Satellite 310 Dead Reckoning Propagation Processor 312 Wheel Step 314 Mode Selector 316 Combiner 328 DR Calibration

Claims (10)

A GPS receiver mounted on a vehicle and subject to signal transmission disturbances from orbiting navigation satellites;
A DR computer coupled to the GPS receiver and providing range and course information calculated by dead reckoning (DR) during the signal failure, wherein vehicle navigation solutions and assistance are through a screen display to the user Being continuously available,
In a car navigation system comprising a wheel step sensor coupled to a DR computer and providing rotation data relating to rotation of individual wheels supporting the vehicle,
Characterized by an estimate of the true radius of the individual wheels,
The estimate of the true radius of the individual wheels is the range change (ie, delta range) and course change (ie, delta course), the wheel step sensor from the wheel increment sensor to generate information about the operation of the vehicle. Combined with rotation data,
The delta range and delta heading information is propagated by the DR computer from the absolute position information periodically obtained from the GPS receiver, and the dead reckoning result is output alone or in some cases as a vehicle navigation solution. ,
The vehicle navigation solution and assistance is continuously available through a screen display to the user and is provided by the GPS receiver alone, by the DR computer alone, or by the GPS receiver and the DR computer. Car navigation system derived from the complex.   A computer program further comprising: selecting and deriving the vehicle navigation solution from the GPS receiver alone, from a DR computer alone, or from a composite provided by both the GPS receiver and the DR computer. Item 1. The car navigation system according to Item 1. A computer program that operates in one of three modes, including a GPS receiver alone, a DR computer alone, and a combined mode that includes both the GPS receiver and the DR computer;
Operation in the GPS receiver single mode occurs when a GPS fix is available, there is a satellite vehicle available sufficiently consistently, and more than two satellite vehicles are being tracked,
Operation in the DR computer single mode occurs when GPS fix is not available or when two or fewer satellite vehicles are being tracked and calibration is available,
Operation in the combined mode has been verified from inspection that no more than two satellite vehicles are being tracked, or that there are no satellite vehicles available consistently, and that calibration is available. The car navigation system according to claim 1, which occurs in some cases. A programmable permanent stop bit;
A branch of a conditional program that immediately follows activation of the GPS measurement hardware, the conditional branch depends on the status of the programmable permanent stop bit, calibration is available, and When the programmable permanent stop bit is true, transfer control to the DR computer single mode, or if the programmable permanent stop bit is chaotic, a GPS fix is available 4. The car navigation system according to claim 3, further comprising: transferring control to the GPS single mode.   Further comprising automatic processing for continuous calibration of an estimate for the true radius of each of the individual wheels, and absolute position information periodically obtained from the GPS receiver is provided by a wheel step sensor. The car navigation system according to claim 1, wherein the car navigation system is compared with the rotation data. A buffer and a time stamp process for receiving a wheel rotation data packet received from the controller area network (CAN);
The car navigation system according to claim 1, further comprising a dead reckoning Kalman filter for estimating and calibrating the wheel step rotation data from the buffer and time stamp processing.   Further comprising an initialization value that is computed to approximate the true radius of each of the individual wheels and is used to provide an estimate of the delta range, delta course, and delta course total for the dead reckoning solution The car navigation system according to claim 1.   The car navigation system according to claim 1, further comprising speed effect compensation for adjusting an estimated value of the wheel radius. Tightly coupled GPS, dead reckoning and map matching navigation software applications that run as a GPS application on a host computer and are connected to receive measurements from a GPS measurement platform (MP) on a wheeled vehicle A program,
Input processing to collect time, fix trigger, Doppler, and pseudo-range reports from a GPS measurement platform (MP) device mounted on a wheeled vehicle;
An offset calculation process from the GPS time to the host computer time, in which the asynchronous data generated by the vehicle is aligned with the GPS time without this process;
Each new position fix and position (P) based on GPS system time (P), velocity (V), calibration parameters for wheel radius and ratio (B), delta course and delta range (D), and least squares snap to time A shot estimator;
Configured for position (P), velocity (V), calibration parameters for wheel radius and ratio (B), delta course and delta range (D), and time, dead reckoning delta range, delta course, combined mode And a Kalman filter to handle propagation,
Depending on the availability of GPS solution and wheel data calibration, continuous mode selection process that operates in GPS single mode, dead reckoning single mode, or combined mode of both GPS and dead reckoning, and
New fix calculation processing that calculates hardware advance data and adjusts the GPS millisecond timer connected to the time offset calculation processing;
Buffer processing to collect time-stamped CANbus data for wheel rotation ticks measured by vehicle wheel transducers;
A _L and A _R calibration parameters for delta course / tick difference between left and right wheels, and B _L and B _R calibration parameters for delta range / tick of left and right wheels, and total delta course as known course Dead reckoning calibration Kalman filter that processes the course offset calibration parameter that translates to the reference,
Dead reckoning data processing including dead reckoning error model, converting the data from the buffer processing into delta range, delta course, and delta course total information for later processing by the Kalman filter;
Application program with application programming interface (API) that outputs new navigation fix A mode selection program that operates in one of three modes including a GPS receiver alone, a DR computer alone, and a combined mode that includes both the GPS receiver and the DR computer;
Operation in the GPS receiver single mode occurs when a GPS fix is available, there is a satellite vehicle available sufficiently consistently, and more than two satellite vehicles are being tracked,
Operation in the DR computer single mode occurs when GPS fix is not available or when two or fewer satellite vehicles are being tracked and calibration is available,
Operation in the combined mode has been verified from inspection that no more than two satellite vehicles are being tracked, or that there are no satellite vehicles available consistently, and that calibration is available. Occurs when
10. The tightly coupled GPS and dead reckoning software application program of claim 9, wherein a new navigation fix is subsequently output through the API.

Images