| Navigation | |
A sensor architecture for high precision UAS navigation
This paper presents the CIRA’s flying test facility for autonomous mid-air flight and landing on runways instrumented by Differential Global Positioning System (DGPS) base station. We present this paper in two parts. The first part focuses on system architecture and navigation algorithm. We would publish the second part of the paper in May 09 issue of Coordinates. The second part will detail the algorithm validation procedure and results. |
|
The effort for realizing fully autonomous and operative Unmanned Aerial Systems (UAS) induced the need of developing innovative techniques for integrating measurements derived from different aircraft navigation systems. Since no human aid is available onboard UAVs, navigation hardware must attain larger capabilities than the ones of manned platforms. In particular, the most important features that shall be considered are autonomy, safety, compatibility with previously developed flight standards, and whole mission coverage [1]. This latter feature means that adequate and reliable navigation must be accomplished even during critical phases such as takeoff and landing, during which the required positioning should be always available with a high accuracy [2].
Several research activities have been developed in last years in order to increase the autonomy features in UAVs [3], in order to expand the flight envelope [4] and to improve security levels of modern aircrafts, both manned and unmanned. In particular, a very great research effort has been devoted to the achievement of high automation in the landing phase, so as to allow the landing of an aircraft without human intervention, also in presence of environmental disturbances.
Currently, no single sensor is capable of reliably realizing the required performance without relying on some ground measurement, hence UAS navigation requirements can be fulfilled only by integration of measurements from multiple sensors. In particular, configurations that integrate inertial sensor measurements with GPS, altimeters, air data sensors, and magnetometers are very frequent ([5]-[7]), and resulting performance and reliability depends on both sensors accuracy and adopted integration techniques.
In this context, CIRA developed a complete autonomous mid-air flight [8], collision avoidance [9], take-off and landing ([10], [11]) system for fixed wing aircrafts. This overall system has been worked out in the national founded project TECVOL (Technologies for Autonomous Flight), which continues and extends the previous CIRA project ATOL (Automatic Take-Off and Landing), successfully completed in 2004 [12]. Autonomous mid-air flight and autonomous landing capabilities of this system have been completely tested, up to the experimental in flight validation by means of a prototypal flight test bed, while at present autonomous collision avoidance and autonomous take off systems have been tested up to the real time stage and are ready for the real world in flight validation.
To work with high performances, the autonomous flight algorithms developed in the TECVOL project need of a good set of navigation measures, so suitable sensors fusion algorithm has been developed in the same project, as described in the next.
System Architecture
The technologies developed in the TECVOL project are validated in real world by means of a flying test bed, consisting of a piloted aircraft equipped with on-board avionic system and onground control station, both designed and integrated by CIRA. The flyingplatform is an experimental ultra light aircraft with a max take off weight of 450 Kg, a max speed s/l about 218 km/h, a cruising speed about 190 km/h, wing area of 13.2 m2, wing span of 9.6 m and maximum engine power of 100 hp.
On board avionic system includes all devices, selected among the Commercial-Off-The-Shelf (COTS) ones, needed to perform the in flight experimental validation of advanced guidance, navigation and control functionalities. These devices are the ones described in the next.
• A Flight Control Computer (FCC), which is an embedded real-time computer based on a PowerPC processor, provided by the supplier with a tool based on the most advanced control system rapid prototyping methodologies. This software environment enables the automatic coding of the real-time SW directly from Matlab/SimulinkTM diagrams.
• A navigation sensor suite including: (a) two GPS, one in differential mode and the other in Real Time Kinematic (RTK) L1/L2 mode, capable to provide position measurements with few centimetres accuracy; (b) a solid state Attitude Heading Reference System (AHRS); (c) a laser altimeter; (d) an Air Data Computer (ADC) with a dedicated air data probe capable to provide IAS, TAS, barometric pressure and aerodynamic angles measurements. For what concerns the two GPS installed on board, it is worth to emphasize that only the one in DGPS mode is used in the sensor fusion algorithm described in the next, while the other, in RTK mode, is used as reference to verify the sensor fusion algorithm accuracy.
• Digital electromechanical servos, to command both aerodynamic surfaces and throttle, driven by the FCC.
• A digital bidirectional data link system able to exchange data between on board FCC and the ground control station, with a maximum bit-rate of 9.600 bit/ sec in uplink and 115.200 bit/ sec in downlink, up to about 6 Km of operative range.
• Angles of attack and sideslip sensors.
• Surfaces position sensors.
As already said, two GPS are installed on board and configured in different ways, to have a comparison between the aircraft position and velocity measures obtained by the sensors fusion algorithm implemented in the FCC and the ones obtained by the GPS configured in RTK mode. The radiomodem output is connected to both GPS, in such a way as each sensor receives and selects the Real Time Correction Messages (RTCM) required by the RTK and DGPS algorithms. The ground control station (GCS) is installed in a big shelter fixed on the ground near the runway. It is designed to show telemetry data to flight test engineers through dedicated
Human Machine Interfaces (HMI) and to allow the on board avionic system remote reconfiguration. The GCS architecture is based on the interconnection of the elements listed in the next:
• Ground Control Computer (GCC),
• downlink radio-modem,
• uplink radio-modem,
• engineering workstation computer,
• virtual cockpit computer,
• autonomous flight management computer,
• GPS base station,
• GPS differential correction radio-modem.
The GCC, based on PC-104 formfactor, is the core of the whole ground segment. It is connected, by an Ethernet field bus, with all the computers in the GCS and, by dedicated point-to-point RS232 connection, with the uplink and downlink radio-modem. Considering reliability constraints, both GCC and FCC use a hard real-time operating system.
Navigation Algorithm
The most common integrated navigation techniques make use of Kalman filtering ([13]-[15]). The main drawback of these techniques is the necessity of an
accurate sensor error model, so as when poor information are available about the sensors used it is very difficult to obtain an appropriate adjustment of the Kalman Filter. With the aim to develop a low cost navigation sensor suite, our choice has been to use a simpler sensor fusion algorithm based on the concept of Complementary Filtering that will be described in the next.
For what concerns, generally, a sensor fusion algorithm design, in a navigation system based only on GPS measures, typical error sources are:
• excessive noise,
• low updating frequency (generally up to 10 Hz), therefore for dynamics the faster measures turn out little reliable,
• satellite “loss”.
To overcome these limitations, a good solution consists in the integration of the GPS position and speed measures with the ones coming from other sensors: the auxiliary system most commonly used to such aim is an INS sensor.
In Table 1, main features of both GPS and INS are shown and compared.
In general, from the Table 1 analysis, it can be derived that the advantages arising by the INS/GPS integration are:
• advanced accuracy on position and speed,
• limitation of the INS errors by means of the GPS measures,
• possibility of using INS when the GPS is unusable,
• the velocity measure from the INS can help to eliminate GPS jamming problems,
• the position measure from the INS can help to reduce the acquisition time of the satellites (time to hot fix),
• the high INS short-term accuracy can eliminate the cycle-slip problem,
• thanks to GPS, INS can be economic and more compact.
Luca Garbarino, Vittorio Di Vito, Ettore De Lellis, Carmine Marrone, Federico Corraro
|
||||||||||||||||||||||||||||||||||||||||||||||||||
| The complementary filter here proposed, described in the next, is a further method which contributes to integrate position and speed measures, coming from GPS, with accelerations, attitude and orientation measures, coming from an AHRS (Attitude and Heading Reference Systems). In this case, it is not necessary the use of a sophisticated INS with its algorithms for estimating, independently from the GPS, position and speed of the vehicle. This filter aims to determine in the best way the aircraft position and speed, in the NEU reference system, by using both the raw measures from the inertial sensors and the measures supplied by the GPS.
The general concept of the complementary filter is the integration of acceleration measures supplied by the AHRS, in order to obtain position and speed measures affected by lower noise and with a larger band in comparison with GPS measures. However, even if the AHRS measures are little noisy, they are affected from remarkable bias errors, so speed and position calculated only by integration of the accelerations can quickly diverge from the real values. In order to limit the effects due to the bias, therefore, it can be thought to integrate the accelerations and to process them through a high-pass filter, obtaining the medium-high frequency component of the considered signals. The low frequency components can be obtained by a filtering stage of the GPS measures through a low-pass filter. The final estimate of position and speed is equal to the sum of the two components above mentioned. The resulting architecture of the complementary filter we developed is, therefore, the one shown in the schematic representation of Figure 1. It is important to emphasize that, in both velocity and position measures estimation, the high-pass filter applied to AHRS measures and the low-pass filter applied to GPS measures must be “complementary”, in the sense that the sum of the transfer functions of the two filters must be equal to one. This is the reason why the navigation measures integration method here proposed is defined “complementary filter”. The specific cut-off frequencies used in the filters shown in Figure 1 have to be chosen to reach the following two contrasting aims: minimizing the noise power due to the GPS and avoiding the error arising from the integration of the AHRS accelerometers bias. The method above described applies in normal no-failure conditions, where INS and GPS sensors correctly work. However, also in the case of GPS failure it is necessary obtain estimation, even if not optimal, of vehicle navigation data. The strategy adopted in this situation is described in the next. In the case of GPS failure, the basic idea is to replace the GPS measures with the ones provided by a sensor characterized by the same characteristics, even if with lower precisions: in this case ADS, with an appropriate offset adjustment, represents a good solution. Pressure altitude (PALT) is used regarding the vertical position measure, while for the vertical speed is used the PALT RATE measure. Regarding position and velocity in the horizontal plane, instead, ADS does not directly supply such measures, but they can be opportunely obtained. In particular, for the velocity in the horizontal plane estimation the procedure described in the next is used. As long as GPS correctly works, it is continuously performed wind estimation, based on the relation: W = Vin – TAS where W, Vin and TAS represent respectively wind, inertial velocity and true air speed vectors, in the inertial reference frame. When a GPS failure is detected, this wind estimation is frozen and constant wind is considered, so from the TAS measure derived from ADS it is possible to approximately estimate the inertial speed as:
Such components are used in place of GPS velocity measures as inputs in the complementary filter, which supplies in output velocity and position estimation. This idea correctly works when the aircraft is following a trajectory in a midair flight mission. In the case of GPS failure during landing, to obtain a better estimation of the measures of interest, it is also possible to use laser altimeter measures. During landing phase, therefore, PALT and PALT RATE ADS measures are replaced by altitude and vertical speed estimations derived from laser altimeter measures. In this case too, of course, the cut-off filtering frequencies applied on the laser are specifically optimized. For what concerns the use of the navigation measures integration method here proposed in the future Global Navigation Satellite Systems (GNSS) framework, furthermore, it is very relevant to emphasize that the described sensor fusion algorithm can be used in this framework too, by simply replacing the GPS receiver with one able to receive EGNOS (European Geostationary Navigation Overlay Service) and GALILEO signals. Moreover, in the future GNSS framework it will be possible to improve the proposed algorithm, by including new safety features. In particular, the basic idea consists in using the EGNOS performance information (in terms of accuracy, integrity, continuity and availability) to improve the sensor fusion algorithm efficiency and to add an integrated diagnostic function for detecting system failures. Based on this integrated diagnostic function, it will be possible to switch, in case of failure, in an appropriate degraded navigation mode. This will constitute a very relevant enhancement of the proposed navigation system, considering that integrity issues, important in general for many applications, are particularly critical in the aviation field, where vehicles can travel at high speed and can quickly deviate from the flight path. |
||||||||||||||||||||||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||||||||||||||||
|
Pages: 1 2
(No Ratings Yet)
Loading...
