Problem
Motivations
IMU and calibration
GPS algorithm
INS algorithm
Kalman filter
DSP programing
Graphical User Interface
We need a reliable navigation system, with high update rate INS: measure acceleration and angular velocity within the body’s reference frame. Use these data to update the position. GPS: measure the distance of the body from known points of reference. Use that distance and the position of the references. Both systems have their advantages and disadvantages OPTIONS: INS or GPS
Difference between the INS and the GPS estimate used to estimate the measurement error in the IMU and correct INS estimates INS: small errors in acceleration huge errors in position GPS: noisy estimate, low update rate A combination of both systems can solve the problem.
The IMU (Inertial Measurement Unit) measures the
acceleration and angular velocity of our cart.
3 measured outputs : ax, az, ωxz
Due to :
. Quantization
. Biases inherent to IMU implementation
… The data we obtain from the IMU has to be modified.
Need for a calibration of the IMU
Calibration of linear accelerations Where :
. x : true and unknown accelerations
. T : missalignment matrix
. k : scaling factors
. b : biases for each axis
. y : known acceleration outputs
Calibration of the gyroscope Where :
. w : true and unknown angular velocity
. k : scaling factor
. b : bias
. y : known angular velocity output
Calibration based on old data Calibration based on new data Trajectory output for old data calibration Trajectory output for new data calibration
Principle:
Estimate distance of mobile from two or more known positions.
Solve the two-dimensional equation system for the unknown position.
GPS Design: Uses satellites with known ephemerides for reference positions.
Sound GPS Design: We use 2 loudspeakers placed at known positions.
Used for the integration of INS and GPS
Range Computation:
Range ρ = vsound·δT
δT = δφ·1/(2πf)
Phase difference between transmitted and received signal translates to range measurement.
Pseudo Noise sequence is transmitted by the loud speaker as a BFSK modulated signal.
Measure the time δT between transmission and reception for both loudspeakers.
Assumption: The receiver knows when the PN sequence was transmitted.
Synchronization transmitter-receiver
Synchronization is accomplished using training sequences.
Receiver tracks training sequence, and finds time reference.
Fixed Initial Position – the initial distance from each speaker must be known. Time of transmission of first PN sequence noise Training sequence PN seq PN seq PN seq PN seq Initial known Range
Used for the integration of INS and GPS
Finally, we have the position estimates from the Sound GPS.
The position estimates contain a bounded high frequency error.
To smoothen this, we use a Kalman filter.
Two reference frames:
Need for a frame transformation matrix
Room
Moving target Based on Newton’s equations
Adaptive filtering technique used to estimate dynamic processes
Based on state model, with a state variable x(n) and an output y(n) It can be used in
Prediction form
Random walk Extended Kalman Filter GPS filtering
Used for the integration of INS and GPS
Two possible states:
We have a GPS estimate: filter the difference between the GPS and INS position estimates and get a new error estimate
We don’t have a GPS estimate: use the previous error prediction to correct the error and update prediction
Noisy trajectory of INS Estimation of errors
Implemented on Code Composer Studio
Link between the data we receive from the DSP card and our algorithms
To obtain real-time implementation
Plan :
. General implementation scheme
. How to retrieve INS values and what to send to the algorithms ?
. Overwriting and error flags
Used for the integration of INS and GPS
General implementation scheme
Used for the integration of INS and GPS
Comments