Home Analog Devices Feedback Subscribe Archives 简体中文 日本語 | |

Download this article in PDF format. (998 KB) Reduced Integration Time Improves Accuracy in Dead Reckoning Navigation Systems By Ben Wang An automotive
With a digital gyro, the integrated rate can be expressed as the sum of rate samples multiplied by the sampling interval: where n is the number of samples, and τ is the sampling interval. The angular error accumulated over time can be expressed as: where n is the number of samples, and τ is the sampling interval.According to the formula, as the required integration time gets longer, the accumulated error gets bigger, as shown in Figure 2. These rate samples, measured using an evaluation board with the ADXRS810 high-performance angular rate sensor, simulate a DR navigation system with 3300 total rate samples recorded. The blue line shows the gyro rate samples; the red line shows the accumulated angular error. It is obvious that the accumulated angular error increases with time.
_{n}, but today’s digital gyros already have very low rate error specifications. The ADXRS810, for example, features 80 LSB/°/sec sensitivity, ±2°/sec offset, and 0.03°/sec/g shock immunity, leaving limited room for improvement. In addition, the algorithm to compensate e_{n} is complicated. Compared to other applications, such as electronic stability control (ESC), for example, the gyro in a DR navigation system can run for long periods of time, as would be the case when the GPS signal is lost as the vehicle travels through a long tunnel. Longer running time causes the accumulated angular error to be more significant in DR navigation applications. If the integration time could be reduced, it would significantly decrease the accumulated angular error. When the gyro is not rotating, the rate output is small, but nonzero, due to gyro noise. The ADXRS810 achieves very low gyro noise and very high sensitivity, making it easy to filter out noise in the digital domain simply by setting the appropriate threshold. This process is equivalent to low-pass filtering, as the gyro rate noise is at a high frequency compared to the rate output due to rotation. Figure 3 shows the LPF version of Figure 2, where all rate samples less than 1°/s are zeroed and, therefore, ignored when doing rate integration. The remaining integration time, considered effective integration time, is only about 16% of the total integration time. This provides a significant reduction of integration time. As a result, the accumulated angular error is also significantly reduced, as indicated by the red line.
In a practical application, the vehicle steering wheel is normally positioned at zero degrees. Thus, the effective integration time for gyro rates can be reduced by ignoring it, just as was done in the experiment described in Figure 3. Figure 4 shows gyro rate samples from a real in-vehicle test. Traveling through a tunnel for about 180 sec, it requires 180 sec for rate integration. Without the LPF process, the accumulated error over 180 sec can be up to 4°, which is too big to correctly determine the vehicle’s location in the tunnel. By implementing the LPF process with a 0.5°/sec threshold, the effective integration time is reduced to only 84 sec, a reduction of about 53%. The accumulated error drops to about 0.5°, as shown in Figure 5. The LPF threshold can be set to achieve the accuracy required for the specific application.
The ADXRS810 high-performance, low-cost digital gyro uses ADI’s innovative MEMS technology, making it a good choice for vehicle DR navigation applications. Housed in a very small package, it provides low offset, low noise, and high rate sensitivity. Temperature compensated on chip, it eliminates the need for an external temperature sensor and eases the algorithm for temperature compensation. Its high immunity to shock and vibration is very important in automotive applications.
Copyright 1995- Analog Devices, Inc. All rights reserved. |