Special issue: Navigation World 2008 No. 3 – October 21 st , 2008 These features of the propagation mechanisms have been observed by many research ers. The detailed results of wideband indoor measurements show a large dependence of the power delay profile on the room layout, furnishing and type of construction material. It is reported that even small metallic obstacles like window handles result in a reflective and diffractive source for electromagnetic waves. Similarly, it was noticed that rooms that are equipped with a lot of metallic objects like factory plants yield a substantially larger delay spread. Directivity of waves in corridors, as well as preferred directions of incident waves through less absorbing materials like doors or windows compared to walls, has been observed by many measurements. All these effects are easy to explain. But the problem remains, that the large variety of different indoor environments results in a large variety of effects on the signal propagation. Therefore, it is very hard for satellite navigation receivers to cope with these highly variable propagation effects. Combination of personal dead-reckoning and map-matching A particularly powerful auto - nomous combination is the use of inertial step measurements in conjunction with map information. The researchers at DLR have demonstrated that such a system can converge to the correct position after less than a minute of motion. DLR has developed a two-layer real-time sensor fusion platform that operates with a Kalman fil- Pedestrian movement models can be used to predict the human behavior. The aim of such models is to characterise pedestrian movement as closely to real life as possible. A movement model that truly imitates the real pedestrian movement can be used for society planning, evacuation plans, buildings construction, human-like robotic movement and indoor/outdoor navigation. Human movement can be described in terms of physical parameters like speed, direction and as a result the position. However, speed and direction are affected by several human states. For example, a pedestrian trying to catch a train is faster than a pedestrian who is shopping. Examples of other parameters that affect pedestrian movement are age, activeness, arousal and emotions. Some of these parameters affect movement more than the others. Building layout also very much affects the movement of the pedestrian. It is clear that a pedestrian can not walk through a wall. Movement constraints that control these physical parameters are categorised into two groups. The first category includes parameters that can be determined accurately, such as age, weather, time of day and weekday, as well as parameters that can be derived from external data, such as ground steepness or obstacles at the ter for the stride estimation, and which fuses other sensors and maps at a higher-level, lowerrate, particle filter. The platform is computationally efficient and flexible enough to join data from a variety of heterogeneous navigation sensors. In buildings, a few dispersed RFID tags or even moderately GNSS reception can significantly aid the overall positioning. Business and Innovation Magazine news for Satellite Navigation 7 Low-cost inertial sensors for personal dead-reckoning The use of fully self-contained inertial sensors is promising for pedestrian navigation, in particular for indoor applications. Two basic approaches can be distinguished. The pedometer approach employs an accelerometer for detecting individual steps whilst the stride length and stride direction are themselves estimated using additional sensors, such as GNSS, or a priori information. Given a detected step, its length and its direction, a person’s posi- Stride ● Rest phase The role of movement models on predicting human behaviour pedestrian’s position. The second category encompasses parameters that vary according to human behaviour. Examples are activity, disorientation, activeness, arousal and emotions. The movement model is a prerequisite in the Bayesian algorithms that are applied in DLR’s sensor fusion algorithms. The prediction stage depends entirely on the movement model to predict where the pedestrian will most likely be at the next time step. With a good movement model, one can predict accurately the position of the pedestrian at the next time step. This will decrease dependency measurements for estimation. It will also allow the system to take correct decisions regarding the position of the pedestrian at the prediction stage, since we will have less contradiction between the prediction and measurements later on. According to the application, different implementation types of movement models may be used. For example, one may use a statistical model for people who are shopping. On the other hand, a more deterministic model is adequate for navigating people. Additionally, special movement actions have to be considered for specific applications. For example, a movement model for firefighters or rescuers has to consider special actions such as sliding, jumping and rolling. tion can be determined by deadreckoning. The latest approaches are based on full inertial navigation with six degrees of freedom. miniature foot-mounted inertial platform comprising triads of accelerometers and gyroscopes is used to dead reckon via a conventional inertial navigation algo - rithm. Rest phases of the foot, which are detected from the accelero - meter signals, trigger zero-velocity (virtual) measurements that are used to update an adjoint Application constraints Today, it is a common understanding in the scientific community that global navigation satellite systems (GNSS) alone are not sufficient to provide accurate indoor localisation. Other techniques or additional sensors should be considered. There are a variety of alternatives, each with different features in terms of costs, infrastructure requirements, accuracy, volume, stability and privacy constraints. Myriad applications also exist, ranging from warehouse logistics to emergence services and from professional to mass-market Kalman filter (ZUPT). Due to the regular ZUPT measurements it is possible to estimate and correct the drift errors, which accumulate in the inertial navigation solution. It was shown that this approach can achieve a very good performance of only 1% to 3% of travelled distance even with today’s lowcost micro-electro-mechanical sensors (MEMS) because the ZUPTs are commonly so frequent that errors build up only slowly during each step the pedestrian makes. Dead-reckoning performance when using foot-mounted MEMS inertial sensors at the Institute of Communications and Navigation of DLR. DLR’s prototype of a shoe with builtin MEMS inertial sensors applications. Each application has its own constraints. No single solution is expected to fit all these constraints. For example, an emergency service can provide the floor of a person in trouble and the room; the user’s device will be low in cost; infrastructure to support localisation cannot be used; the service needs to be globally operational. On the other hand, in industrial applications, robots or goods will be located with sub-dm precision; the technical solution can be more expensive, proprietary and local; infrastructure is feasible. The technical solutions encompass the use of pseudolites, ultra Combining complementary sensors GNSS: 3D-PVT; Mobile communications: 2D-P Long-term stability, short-term errors INS: Orientation change (3D), velocity change (3D); very short-term stability, long-term errors Optimal sensor fusion algorithm Barometer: Altitude change; short- to mid-term stability, long-term errors Future indoor applications Once the indoor navigation problem is solved technically, a plethora of novel indoor applications becomes feasible: Precise monitoring and guidance of rescue teams, firefighters, and other relief units as well as, amongst many others, interactive museum guides, personal travel assistance, sports applications, and assisted living for elder people. The WayWatch is DLR’s application target for personalised navigation. It will show not just time, but also the way to go in form of picto - grams showing the next steps on an itinerary, such as how to get to the right departure gate, to a certain office room, or to a product in a depart ment store. wide band (UWB), the use of mobile radio signals, RFID, Bluetooth, and WLAN together with fingerprinting. Amongst others dedicated indoor solutions are the systems Active Badge, Cricket, LEASE, Topaz, and the Thinking Carpet. The Institute of Communications and Navigation of DLR focuses its indoor navigation activities on techniques for applications where people are on the move. DLR therefore investigates promising approaches that apply several sensors together with GNSS: inertial systems (INS), compasses, RFID tags, and alti - meters; furthermore, appropriate movement models and maps are of great importance. Positioning in buildings and other environments will not only require a combination of sensors, but also other information such as building plans in order to provide a high level of accuracy. Sensor fusion approaches that combine several complementary sensors such as GNSS, mobile radio positioning, WLAN, RFID, foot-mounted inertial sensors, electronic compasses, baro-altimeters, and maps are promising candidates to solve the indoor positioning problem. A central role here is played by the development of optimal sensor fusion algorithms, which make a joint system superior to the sum of its components due to synergetic effects arising from the complementary properties of the respective sensors and subsystems. Long-term accuracy Personal navigation sensors and their properties RFID, WLAN: 3D-P Long-term stability, short-term errors Step counter INS: Estimation of step sequence and length: Long-term stability INS / measurement Foot-mounted INS: With Zero Update (ZUPT). Velocity: Long-term stable, position error linear in time 3D Pedestrian Movement Model: Some short-term assistance 3D maps (map matching): Long and short-term assistance in buildings Magnetometer: Orientation (2D or 3D); long-term stability, short-term errors Maps and movement Attitude / orientation
Sie befinden sich auf der Textversion einer Flash-Seite.
Um die Seite in ihrem vollem Umfang betrachten zu können, benötigen Sie den Flash Player ab Version 8, sowie aktiviertes Javascript.
Sie können jederzeit versuchen, diese Abfrage zu überspringen und das Magazin trotzdem aufzurufen, indem Sie hier klicken!