**2.3. Demonstrator**

To prove robustness and applicability, research results obtained within CoLOR were experimentally validated in an extensive measurement campaign. The main focus was placed on a robust transfer of the algorithms from laboratory conditions to a more realistic indoor propagation scenario. The goals of these extensive multi-disciplinary measurements were

• to verify UWB sensing algorithms in real indoor scenarios,

Most algorithms of previous investigations could be adopted with manageable complexity to the mobile bat-type scenario. However, some challenges came up unexpectedly, e.g. erroneous robot motion due to uneven floor and slippage of robot tires, potentially more clutter in a realistic scenario, handling and transformation between a global coordinate system of the static environment and an additional local coordinate system of the dynamic robot. To provide a realistic indoor scenario with corners as well as edges and dimensions like that of a larger office room (56 m2), the fire detection laboratory of the University Duisburg-Essen was modified and used as the location for the measurement campaign. The modification consists of partly installed portable metallic walls to give the room a more complicated shape.

Cooperative Localization and Object Recognition in Autonomous UWB Sensor Networks 185

The ground plane of the designed indoor scenario is depicted in Fig. 4 .

9.24 m

possible track

o4

2.59 m

o2 o3

5.6 m

As previously mentioned, the bat-type sensor is realized by a mobile security robot which is

As super-resolution techniques in short-range UWB sensing have to be performed, certain demands on the motion accuracy result are necessary. However, localization accuracy is predominantly achieved by advanced algorithms described later in chapter 7. Additionally, some assisting robot specifications have also been taken into account to improve accuracy. There are three actuators in the robot, two in the motion unit at the bottom, and one at the top which rotates the antenna array. The actuators are all hollow-shaft servo motors, which offer unique features unsurpassed by conventionally geared drives. Used in highly demanding industrial and medical servo systems, they provide outstanding precision motion

o1

0°

90°

270°

objects (not true to scale)

180°

7.91 m

o5

schematically represented in Fig. 5

robot

3.9 m 2.75 m

**Figure 4.** Ground plane of measurement location and robot with possible track.

**Figure 3.** Simulated scenario.


With regard to a final experimental validation, the investigations in CoLOR were accompanied by experimental practice throughout the project, i.e. algorithms were always validated with regard to real world conditions.

The measurements made so far were simplified to limited sensor and/or object motions, to ideal movements by means of motorized linear arrays or by reducing the number of simultaneously performing cooperative algorithms. Therewith, only hardware complexity and mechanical effort could be reduced and meaningful validations of the investigated algorithms could be obtained. However, measurement scenarios and instrumental investments were extensively expanded within the scope of a demonstrator to fully meet the demands of a realistic indoor propagating scenario. Hence, an autonomous mobile security robot with professional motion units was fully equipped with UWB-devices, RF components, a power supply unit and a laptop for data acquisition and communication with the data fusion computer. This security robot serves as the previously mentioned mobile bat-type sensor.

Most algorithms of previous investigations could be adopted with manageable complexity to the mobile bat-type scenario. However, some challenges came up unexpectedly, e.g. erroneous robot motion due to uneven floor and slippage of robot tires, potentially more clutter in a realistic scenario, handling and transformation between a global coordinate system of the static environment and an additional local coordinate system of the dynamic robot.

6 Will-be-set-by-IN-TECH

• to perform and evaluate the simultaneous working algorithm for the cooperative

• to determine to what extent previously achieved research results are applicable under these

With regard to a final experimental validation, the investigations in CoLOR were accompanied by experimental practice throughout the project, i.e. algorithms were always validated with

The measurements made so far were simplified to limited sensor and/or object motions, to ideal movements by means of motorized linear arrays or by reducing the number of simultaneously performing cooperative algorithms. Therewith, only hardware complexity and mechanical effort could be reduced and meaningful validations of the investigated algorithms could be obtained. However, measurement scenarios and instrumental investments were extensively expanded within the scope of a demonstrator to fully meet the demands of a realistic indoor propagating scenario. Hence, an autonomous mobile security robot with professional motion units was fully equipped with UWB-devices, RF components, a power supply unit and a laptop for data acquisition and communication with the data fusion computer. This security robot serves as the previously mentioned mobile bat-type sensor.

**Figure 3.** Simulated scenario.

regard to real world conditions.

approach,

conditions.

To provide a realistic indoor scenario with corners as well as edges and dimensions like that of a larger office room (56 m2), the fire detection laboratory of the University Duisburg-Essen was modified and used as the location for the measurement campaign. The modification consists of partly installed portable metallic walls to give the room a more complicated shape. The ground plane of the designed indoor scenario is depicted in Fig. 4 .

**Figure 4.** Ground plane of measurement location and robot with possible track.

As previously mentioned, the bat-type sensor is realized by a mobile security robot which is schematically represented in Fig. 5

As super-resolution techniques in short-range UWB sensing have to be performed, certain demands on the motion accuracy result are necessary. However, localization accuracy is predominantly achieved by advanced algorithms described later in chapter 7. Additionally, some assisting robot specifications have also been taken into account to improve accuracy. There are three actuators in the robot, two in the motion unit at the bottom, and one at the top which rotates the antenna array. The actuators are all hollow-shaft servo motors, which offer unique features unsurpassed by conventionally geared drives. Used in highly demanding industrial and medical servo systems, they provide outstanding precision motion

end-fire radiating antenna (*Tx*1\_1/1\_2*Rx*1\_1/1\_2). The requirements as well as the design of the

Cooperative Localization and Object Recognition in Autonomous UWB Sensor Networks 187

An array of switches, as shown in Fig. 6, allows the change between the different receiver and

In the framework of this project, a realistic UWB multi-path propagation simulation tool was developed in order to test and compare different algorithms and antenna arrangements for indoor UWB sensing and imaging. Multi-path propagation implies that the transmitted signal does not only arrive over the direct propagation path at the receiver, but also over paths which are dependent on the propagation environment in a complex manner. The received signal is then a combination of a multiplicity of reflected, diffracted and scattered electromagnetic waves. Wave propagation models, in general, can be classified into deterministic and stochastic ones. Deterministic models are based on the physical propagation characteristics of electromagnetic waves in a model of the propagation scenario. In contrast, stochastic models

By now, some statistical channel models have been established for the early design phase and for testing the ideas for possible applications. Statistical models randomly generate channel impulse responses of a channel based on the probability functions, which are usually obtained from measurements. However, if a system has to be tested in a specific environment, deterministic channel models are required, which approximate real physical phenomena. One of the most popular deterministic channel modeling approaches is ray tracing (RT), based on geometrical optics and the uniform theory of diffraction. In outdoor areas ray tracing simulations emulate the propagation conditions very well [18]. Furthermore, it has been shown that ray tracing can be also easily extended to simulate ultra-wideband channels.

However, comparisons between the measurements and simulations with respect to UWB indoor channels show that the ray tracing results are often underestimated in terms of received power, mean delay and delay spread [26, 34, 42]. This is due to insufficient modeling or the

Diffuse scattering causes contributions to the power delay profile, which are not resolvable (dense multipath components). Through these contributions the power delay profile is

antennas itself are described in detail in subsections 7 and 6 .

**Figure 6.** Schematic diagram (left) and picture (right) of the antenna array used

**3. Hybrid deterministic-stochastic channel simulation**

describe the behavior of the channel through stochastic processes.

complete neglect of diffuse scattering in the ray tracing model.

smoother than in the scenarios with reflection and diffraction only.

transmitter configurations.

**Figure 5.** Schematic drawing of the motion unit (left) and of the robot (right).

control in sub-mm range and high torque capacity in a very compact package. The robot has 3 solid rubber tires on both sides which are connected by a chain-drive. Actually, both sides could be used autonomously with different acceleration, deceleration and speed, which results in curved tracks. However, for further accuracy different triggering is avoided so that the servo motors are driven synchronously. Evidently, the robot moves straight forward when the motors drive into the same direction and rotation is performed when the motors drive in the opposite direction. To maintain a more gliding rotation of the robot with reduced positioning errors, the circumference of the middle tire is minimally higher than those of the other ones. The dimensions of the robot as well as the tire position maintain a rotation center in the middle of the robot which also equals the middle of the bat antenna platform at the top. Hence, the movement of the robot was entirely restricted to translations and rotations, strictly avoiding curvature paths. Because of that, a track is split into several straight segments which are separated by a change of orientation. A resulting possible track is shown in Fig. 4. As mentioned previously, to further minimize erroneous robot motions, the bat is equipped with its own rotational unit. The orientation change can be performed by just rotating the bat, which is preferred compared to rotating the whole robot. This is more sensitive to errors due to an uneven or slippery floor. The robot is also equipped with a laser-based indoor navigation system. This highly accurate and well-proven localization system shall provide reference data for subsequent performance analysis. It has neither assisting nor guiding functionality in the localization process of CoLOR. The localization process is first and foremost handled by UWB-Radar technology.

Due to the different localization and imaging applications in this project, the requirements placed on the antenna characteristics differ. Fig. 6 gives an image and a photograph of the antenna array used. It consists of three different antenna types. A broadband monopole antenna (*Tx*2\_1), a dual-polarized broadside radiating antenna (*Tx*2\_2, *Rx*2\_1, *Rx*2\_2) and an end-fire radiating antenna (*Tx*1\_1/1\_2*Rx*1\_1/1\_2). The requirements as well as the design of the antennas itself are described in detail in subsections 7 and 6 .

**Figure 6.** Schematic diagram (left) and picture (right) of the antenna array used

8 Will-be-set-by-IN-TECH

**Figure 5.** Schematic drawing of the motion unit (left) and of the robot (right).

UWB-Radar technology.

control in sub-mm range and high torque capacity in a very compact package. The robot has 3 solid rubber tires on both sides which are connected by a chain-drive. Actually, both sides could be used autonomously with different acceleration, deceleration and speed, which results in curved tracks. However, for further accuracy different triggering is avoided so that the servo motors are driven synchronously. Evidently, the robot moves straight forward when the motors drive into the same direction and rotation is performed when the motors drive in the opposite direction. To maintain a more gliding rotation of the robot with reduced positioning errors, the circumference of the middle tire is minimally higher than those of the other ones. The dimensions of the robot as well as the tire position maintain a rotation center in the middle of the robot which also equals the middle of the bat antenna platform at the top. Hence, the movement of the robot was entirely restricted to translations and rotations, strictly avoiding curvature paths. Because of that, a track is split into several straight segments which are separated by a change of orientation. A resulting possible track is shown in Fig. 4. As mentioned previously, to further minimize erroneous robot motions, the bat is equipped with its own rotational unit. The orientation change can be performed by just rotating the bat, which is preferred compared to rotating the whole robot. This is more sensitive to errors due to an uneven or slippery floor. The robot is also equipped with a laser-based indoor navigation system. This highly accurate and well-proven localization system shall provide reference data for subsequent performance analysis. It has neither assisting nor guiding functionality in the localization process of CoLOR. The localization process is first and foremost handled by

Due to the different localization and imaging applications in this project, the requirements placed on the antenna characteristics differ. Fig. 6 gives an image and a photograph of the antenna array used. It consists of three different antenna types. A broadband monopole antenna (*Tx*2\_1), a dual-polarized broadside radiating antenna (*Tx*2\_2, *Rx*2\_1, *Rx*2\_2) and an An array of switches, as shown in Fig. 6, allows the change between the different receiver and transmitter configurations.
