Navigating the City

Pages | 1 | 2 | 3 | 4

Proceed on Foot To make this a complete city navigation system, imagine that we take the system to our car. While turning on the motor (the car is still motionless), GPS provides the initial position. We then turn GPS off and drive in heavy traffic for 20 minutes. At that point, we still have an accurate position estimate even though GPS has been off for 20 minutes - although if there is too long a period without a stop, GPS must be turned on again and the INS re-initialized. At the end of the drive, we park, leave the car and turn GPS on again to get a position fix. The program now recognizes that the user is walking and applies algorithms for pedestrian navigation.

Operation of the pedestrian navigator can be divided into three parts:

• step detection
• step length estimation
• and heading determination.

Using physiological models, advanced algorithms, and available MEMS sensor technology, we designed a pedestrian dead reckoning (PDR) navigator. This PDR can also receive assistance from low-power sensors such as a barometric altimeter and a magnetic compass.

In this phase we constructed a handheld pedestrian navigator capable of providing position estimates when GPS is not available, or to reduce GPS usage for power consumption reasons. The main challenge is to maintain position accuracy when no satellite data is available, in urban canyons or inside buildings.

PDR Module Because there are as yet no commercially-available IMUs small enough to fit into handheld devices, we constructed one. We chose the sensors based on availability, performance parameters, cost, and size. We integrated a sensor assembly of three MEMS gyroscopes and accelerometers capable of 1 milliradians/ second and 0.7 milligravity rms noise levels, respectively, with a GPS receiver to form a handheld unit. The unit also includes a 3-D magnetic compass and barometric altimeter. The small form factor allows system testing in real-world situations, giving better knowledge of the actual environment and pedestrian movements.

Different IMU placements on the human body - in a pocket, in a backpack, or handheld while walking - cause large variations in the sensor signals. Therefore, old approaches to pedestrian navigation - systems more wearable than portable, using fixed sensors of one or two accelerometers attached to the user in fixed orientations, measuring vertical and horizontal accelerations - are not entirely valid. The new mode, where sensor orientations may change, requires new algorithms to give accurate solutions.

Step Detection Several methods can determine when a footfall occurs. One such method detects peaks of vertical acceleration corresponding to step occurrences.

Figure 7: Typical acceleration measurements over four sequential steps.

In this study, footfall recognition relies upon determination of the zero-crossing frequency from the acceleration data. Using three orthogonal accelerometers to measure pedestrian movements, we can extract the magnitude of the acceleration vector from one footfall to another regardless of the local gravity component. IMU orientation does not affect this acceleration measurement.

Step detection is triggered when zero-crossing fits within the recognition window. Figure 7 shows a typical acceleration pattern over four successive footfalls, and the instants when step detection is triggered. Use of a window rules out step indications too far or too close to each other with respect to time. If the step time exceeds the window, the step is considered not taken. This method gives robust step detection with minimal computational cost. Step frequency can be evaluated using time difference between sequential footfalls.

Step Length Estimation The key point of a dead reckoning algorithm for a pedestrian navigation system is user step length. Step length can be defined as the distance between the sequential heel impacts, and depends on several factors, mainly velocity and step frequency.

оффшорные зоны . обои для стен . клинкер

Pedestrian dead reckoning (PDR) unit assembled for the tests

The algorithm gives reasonably accurate step length estimates independently of the motion state. Due to the differences in how people adapt their walking under different conditions, calibration of the algorithm can be performed when GPS signals are available.

The system adapts to changes in the walking pace. It can also operate with moderate accuracy even if the initial step length is not known and calibration from GPS is unavailable.

Heading Determination In traditional INS implementation, even small errors in computed IMU orientation cause significant errors to the position estimates, due to misinterpretation of the gravity component in the acceleration measurements. Accurate orientation information is not so critical for the PDR system. This is why heading estimation can be performed using a combination of a magnetic compass and a low-cost MEMS gyroscope triad.

IMU for the pedestrian unit consists of three gyros, three accelerometers, three magnetometers, barometer, GPS receiver, and battery.

This project considered the gyroscopes as the main source of heading information, using the compass as a reference measurement. Local level for the compass heading computation can be obtained using the accelerometers' measurements. A Kalman filter combines the gyroscopes data and compass heading.

In this way, the compass can provide azimuth measurements without time-dependent drift characteristics, enabling determination of the gyroscopes' bias levels. The compass together with accelerometers can also obtain initial IMU orientation. The main drawback of using a magnetic compass is the unpredictable disturbances. Especially indoors, these perturbations can render the magnetic heading information useless in some cases.

To cope with this problem, gyroscopes can detect and compensate for magnetic perturbations in the compass measurements, through comparison of the heading and heading rate of change information between compass and gyroscopes. When these corrections are applied, the indoor navigation solution improves significantly.

Because the IMU has not been rigidly attached to the pedestrian, the compass or gyroscopes heading is not necessarily the walking direction. Before using pedestrian navigation algorithms, the difference between the actual walking direction and IMU-measured heading must be solved. Even if the initial heading and position is not known, the navigator can provide information about the traveled distance and walking velocity. When GPS is available, heading difference and position can be solved and the traveled path corrected.

Pages | 1 | 2 | 3 | 4