Accurate localization of a remote vehicle

Question

I've been working on a project where I'm supposed to design a vehicle that carries antennas to perform SAR (Synthetic-aperture radar) imaging. This vehicle has 3 wheels which 2 of them steered by a DC motor. In order to accurately perform SAR imaging algorithm, I need to know how much my car moved from its last location. There are several methods I've looked up so far, However there are problems with each of that:

• GPS: Using GPS, I can do some calculations to see how much my car moved. However, since GPS locations are off about 4-5 meters, calculations that I'll make won't be accurate at all.

• Ultrasonic Sensor: The problem about this way is limited distance range. (As I know 5-10 meters)

• Accelerometer: Even though it seems mathematically possible to calculate distance traveled using instant acceleration, the errors occurring over time makes this calculation meaningless. There are some proposed ways like Kalman filtering to increase accuracy of these calculations.

• Computer Vision: Using CV, I can track something on the vehicle and calculate its distance from the camera. I don't have much experience about this, therefore I can't estimate its accuracy.

• Encoder: This way consists an encoder to track how much the wheels have turned. This seems to be a good way however I don't know if I can be accurate with this method.

I'm aiming for at least ±5cm accuracy. The total distance may reach dozens of meters.

In above ways or others, what would be a good possible solution to overcome this?

Answer 1

1) Calibrated Odometer method with GPS sync

Car Speedometers and odometers are often inaccurate due to choices with oversized tires. But with on-board computers and some calibration routine , it is theoretically possible to get high accuracy. The integration of this error must be specified by your system specs.

You must start with an error budget and allocate in parts per million xxx ppm for each source of error: mfg tolerance , tire, wear , cornering slip, loose surface grip, brake skids, elevation changes in Z axis.

Then determine when resync is needed with GPS. Keeping a rolling history of cumulative Position Error , forever, for create an algorithm for forward error correction from each of these above error budget sources. Then create Forward Predictor Correction factors for each cause to reduce the need to resync with GPS.

It may depend on your accuracy tolerance, and integration time response specs to determine the optimal trade off between quadrature dual wheel incremental encoders and GPS . Compute the sensitivity to errors if you wish to find the opportunities for improvement.

Always start with good specs with assumptions then verify, improve and repeat until ok.

2) Alternatively, if you rely on a mobile with MEMs accelerometer, Gyro and GPS, and just use an SDK to interface to the API , then it becomes a selection process and software needs starting again with a MODIFIED spec.