WO2006075158A1

WO2006075158A1 - Evaluation of the performance of systems

Info

Publication number: WO2006075158A1
Application number: PCT/GB2006/000099
Authority: WO
Inventors: Marko Bacic; Ronald Daniel
Original assignee: Isis Innovation Limited
Priority date: 2005-01-11
Filing date: 2006-01-11
Publication date: 2006-07-20
Also published as: GB0500502D0

Abstract

Simulation apparatus for evaluation of a system's (1) , such as a vehicle's, performance is described. The apparatus comprises imposition means (2), simulation control means (13) and sensor means (3), such that, in use, the system is arranged to be in communication with the imposition means and with the sensor means. The simulation control means is operative to cause the imposition means to act on the system in response to a signal received from the sensor means so as to provide a simulated effect corresponding to said signal. The imposition means may comprise a moveable vehicle support structure. The apparatus may additionally comprise controllable fluid flow means (15) , such that, in use, the vehicle is positioned in the path of the fluid flow. Also described is apparatus for controlling a production process, comprising a plurality of spaced apart sensors.

Description

EVALUATION OF THE PERFORMANCE OF SYSTEMS

The present invention relates generally to simulation apparatus and method for evaluating the performance of a vehicle or system. In particular, but not exclusively, the present invention relates to simulation apparatus for evaluating the control systems of unmanned air vehicles (UAVs) . It should be appreciated however that the present invention finds utility in evaluating the performance of many other types of vehicle including other types of aircraft, marine craft, land vehicles and spacecraft. The present invention can also be used where validation is required of any system whose behaviour is dependent on contact with other systems, or in general, an environment. In another aspect of the invention there is provided apparatus and method for controlling a production process.

Control system validation is an expensive and time-consuming task that forms a key stage in the development of any system. The present invention seeks to provide methodology and apparatus applicable to the design and validation of control systems for systems whose behaviour is dependent on its physical environment, and in particular, unmanned drones.

Current control system technology involves non-linear design followed by extensive trials in which the early stages are fraught with design risk. For manned aircraft, extensive simulation is followed by test-flights by specialist pilots. For non-safety critical systems, simulation using a mathematical model can be considered adequate. There is no such option for UAVs, whereas for expensive vehicles, such as missiles, partial hardware-in-the-loop systems have been developed for verification of the guidance system. UAVs are subject to tight performance constraints and need minimal operator intervention and there is at present no clearly identified methodology for their design and, in particular, for the design of their controllers. There is also a need for flight clearance before any such UAV would be allowed to be used in civilian airspace. This can only be obtained if there is a clear demonstration of the viability and safety of any control system used for the UAV.

Control system design requires the use of a system model - or at least a parameterisation of a candidate system model in the case of an adaptive controller. Such system models can be either explicit or implicit. The performance of any controller may thus depend very much on the quality of the model used during controller design.

System models can be obtained through identification of the physical system directly, or from the underlying physics, or through the combination of both, either on-line or off-line. However, the resulting model will never quite capture all the characteristics intrinsic to the original system. To ensure the stability of a real plant, control design is carried out conservatively, often resulting in significant sub-optimality in order to achieve robustness to modelling errors. Even given such conservative design, there is no guarantee that all the quirks of the real system have been taken into account.

Hardware-in-the-loop (HWIL) simulation offers the possibility of reducing the conservativeness of the controller by including the real system in the simulation loop during controller design. It also offers the capability of including non-modelled aspects of a system's behaviour to the satisfaction of agencies that monitor the certification process. The basic idea is straightforward: rather than testing the control algorithm on the mathematical model from which it is derived, use real hardware in the simulation loop.

According to a first aspect of the invention there is provided a simulation apparatus for evaluation of a system's performance, the apparatus comprising imposition means which is adapted to be arranged in communication with the system, simulation control means and sensor means, the apparatus being such that, in use, the system is arranged to be in communication with the imposition means and with the sensor means, and the simulation control means is operative to cause the imposition means to act on the system in response to a signal received from the sensor means so as to provide a simulated effect corresponding to said signal.

The invention may preferably be viewed as simulation apparatus for validating a system's interaction with its environment, the apparatus comprising a motion, force or other dynamic variable imposition means, simulation control means and sensor means, the apparatus being such that, in use, the system's interaction with its environment is transparent.

According to a second aspect of the invention there is provided a method of evaluating a system's performance comprising arranging that the system is in communication with imposition means and with sensor means, the method further comprising causing the imposition means to act on the system in response to a signal received from the sensor means so as to provide a simulated effect corresponding to said signal.

The use of UAVs in civilian airspace is still an open question due to the absence of the human factor on board. If such a vehicle is to be used in civilian airspace, its control system must satisfy numerous regulatory requirements imposed by the civilian authority responsible for the use of the civilian airspace. Certification can only be obtained if there is a clear demonstration of the viability and safety of any control system used for the UAV. We have realised that it would be particularly advantageous to provide a HWIL system for UAV control system that provides a simulated testing environment with a high level of reality. Such a system would advantageously allow for the flight test of a UAV for flight certification where free flight certification flights are impractical.

In a highly preferred embodiment of the invention we propose a wind tunnel-based HWIL simulator for design and validation of control systems. The purpose of the wind tunnel is to add realism - the ultimate goal being that from the UAVs reference frame the simulation is equivalent to the actual flight. Such hardware-in-the-loop simulation requires careful design of the in-flight environment. A tailor-made control algorithm, such as a non-linear model predictive control, is used to ensure the correct wind velocity is maintained together with computer control via a force-feedback active robot wrist. The robot wrist is programmed to respond as though the UAV was flying, the speed of the wind in the tunnel being coupled to the forces being sensed. The inertial forces are sensed by a force-torque sensor and D'Alembert's principle is used to compute the accelerations of the UAV. The information on the UAV acceleration is used to recover the UAVs trajectory using Frenet Frame theory. This information is then used to position the robot wrist in such a way as to track the UAVs motion and move as though in free flight.

Use of a HWIL simulator differs from existing design methods that concentrate on predicting and attaining performance robustly (ie robustness of performance) using sensitivity analysis obtained from a mathematical model. The concept of transparency, borrowed from teleoperation, is used to describe the property required to achieve the new aim of robustness of prediction.

The present invention seeks to validate a system by testing real hardware within a transparent environment. Transparency is a measure of how realistically a human couples to his or her environment through the medium of a remote teleoperation device. The same concept may be applied to physical systems where interaction is through physical or other contact via force, heat, flow etc. A system boundary is defined that separates the system from its environment, which might be a support point, wings, a hull, or other physical component allowing interaction with its environment. Transparency is then a measure of how well the generalised impedances of the boundary so defined are matched in the simulated environment, eg how the force on a wing transports aerodynamic lift to the object's mass, while the object itself behaves as though it is in an environment that is free to react to this force.

If the physical system is unable to distinguish between contact with a real environment and a simulated environment, then the simulator is said to be transparent with respect to this boundary.

The present invention seeks to provide transparency in simulating an object's interaction with its environment.

Of particular interest in the case of the present invention is the incorporation of a real system within a simulated environment to achieve maximum transparency and validity. A valid test of a system is one where all properties of the system under test are present (hence the need for the presence of the real system and not a mathematical model) and the system responds within a maximally transparent simulator. The bounds on the validity of the test can be assessed by the quality of the transparency achieved.

The present invention seeks to provide a maximally valid test of a control system using a simulated environment. In a highly preferred embodiment there is provided a HWIL simulation for testing dynamic objects, such as aircraft, in a wind tunnel. The use of robotic mechanisms with force sensing provides an active flight test platform for real-time testing of air vehicles, particularly small UAVs. This allows for simulation of un- tethered flight within the wind tunnel.

According to a third aspect of the invention there is provided apparatus for controlling a production process comprising a plurality of spaced- apart sensors, the sensors being spaced in a downstream direction of a production source, the sensors being adapted to measure a physical aspect of a substance issued by the production source, the apparatus further comprising production source control means and the arrangement being such that, in use, the production source control means receives signals from the sensors and controls the production source at least in part in response to the received signals.

According to a fourth aspect of the invention there is provided a method of controlling a production process comprising receiving signals from a plurality of sensors, the sensors being spaced in a downstream direction of a production source, the signals being indicative of measures of a physical aspect of a substance issued by the production source, the method further comprising processing the signals and controlling the production source at least in part in response to the received signals.

Various embodiments of the invention will now be described, by way of example only, with reference to the accompanying drawings, in which: Figure 1 is a schematic side elevation of an unmanned air vehicle (UAV) under test mounted on a robotic wrist in a wind tunnel;

Figure 2 is a schematic side elevation of the UAV of Figure 1 (the robotic wrist being omitted for simplicity) in a wind tunnel with a fan and intermediately disposed of the UAV and the fan a plurality of spaced Pitot tubes;

Figure 3 is a block diagram of the various components of Figure 2 together with a simulation control computer;

Figure 4 is a schematic side view of the UAV of Figure 1 mounted on a first two degrees of freedom robotic wrist;

Figure 5 is a schematic side view of the UAV of Figure 1 mounted on a second two degrees of freedom wrist;

Figure 6 is a schematic representation of the transport delay phenomenon;

Figure 7 is a schematic representation of an inventive state estimator arrangement; and

Figure 8 is a schematic representation of Figure 2 showing how the methodology of Figure 7 is implemented to accurately control air speed across the UAV.

Figure 1 shows a schematic side view of an unmanned air vehicle (UAV) 1 of which a control system thereof is to be tested. The UAV generally comprises an airframe, the airframe comprising a fuselage, two wings and a tail fin. The UAV 1 is mounted on a three degree of freedom robotic wrist (shown generally at 2), and the robotic wrist is connected to the UAV 1 by way of a six axis force-torque sensor 3. The robotic wrist 2 is mounted on the floor 9 of a wind tunnel. Attachment means 8 is provided between the UAV 1 and the force-torque sensor 3. The attachment means is sufficiently light and rigid not to impart parasitic dynamics to the system under test.

As shown in Figure 2, towards an opposite end of the wind tunnel there is provided a fan assembly 15. Disposed intermediately of the UAV 1 and robotic wrist 2 and the fan 15 there is provided a plurality of Pitot tubes 16 which are spaced apart in the path of air from the fan assembly to the UAV/robotic wrist combination.

With reference to Figure 3, the UAV/robotic combination, the array of Pitot tubes and the fan are connected to a simulation control computer 13.

The simulator computer 13 comprises data processing means and memory means.

The control system of the UAV comprises an onboard data processor (and associated memory) 10 and inertial navigation sensors 11, the outputs of which are fed to the onboard data processor. The data processor 10 is also connected to an output of a Pitot tube 12 which is adapted to measure air speed and is positioned on or adjacent to the UAV 1. The inertial navigation sensors 11 are of the rate gyro type and are operative to measure the rate of angular change of the UAV in each axis of freedom corresponding to attitude, roll and yaw.

The data processor 10 serves as a guidance control means in that on receiving data from the inertial navigation sensors, the wind speed sensor, satellite navigational data (for example by way of radio frequency link) in conjunction with any predetermined instructions or communicated instructions, the data processor is operative to operate servomotors connected to the various control surfaces (ie ailerons, elevator and rudder) and control the course of the UAV accordingly.

For reasons of simplicity of illustration a robotic wrist having two degrees of freedom will now be described with reference to Figures 4 and 5.

As previously discussed in order for a flight simulation to be valid it has to be transparent with respect to the means of reproducing the frame of reference defining the trajectory of the UAV, the Frenet Frame. This means that the robotic wrist must move without delay or introducing significant phase lags in its motion. However, the support means is typically moving at a very low velocity or is stationary. Geared devices at very low velocities do not transmit forces using simple gear relationships but have non-minimum phase zeros due to complex interactions when nearly stationary. Non-minimum phase zeros are a way of describing the effects of apparent time delays within the gear mechanism because of its complex behaviour in delivering motion at low speed. Non-minimum phase zeros destroy the capacity of the Frenet Frame controller to be transparent, as the object under test will not experience motion of the robotic wrist without a time delay or inappropriate motion.

Gears provide a robust and cheap method of providing drive to the support and can be made to have no non-minimum phase zeros provided they are never stationary. For the gears to be non-stationary while the support is stationary requires redundant drives, where there are more motors than degrees of freedom. If the motors and their gear reductions can be made to act differentially they are then never stationary.

In order for there to be continuous motion for a two degrees of freedom support mechanism there must be at least three motors.

Figure 4 shows one possible embodiment in which two motors are used per axis, each run differentially, which requires four motors Ml , M2, M3 and M4 for the two degrees of rotational freedom.

Each joint axis is shown with two motors. The vertical base axis Y-Y is driven by motors Ml and M2 running in opposition and the horizontal axis X-X is driven by motors M3 and M4 running in opposition. The difference in speed of each individual motor within a pair is controlled to achieve the net motion required of the robotic wrist.

A third axis can be added to provide full control over the attitude of the

UAV.

For N degrees of freedom it is in fact only necessary to have N + 1 motors if all joints are to be driven differentially and thus reduce the number of motors required. An embodiment for two degrees of freedom using three motors is shown in Figure 5. A simple application of the same principle of design would result in a four motor design for three degrees of freedom.

To use N + l motors for N degrees of freedom, a freewheel motor must be provided. Motor M2 acts as a freewheel and carries a horizontal bearing 30 to take the shaft 25 of motor M3. Motor Ml drives a shaft 27 that is connected to a cage 26a. The cage 26a rotates with the output of motor Ml and is connected to a bevel gear 26b. There are thus two outputs from motor Ml , one driving the shaft 27 and one connected to the bevel gear 26b. Thus the shaft 27 and the bevel gear 26b rotate together rigidly.

The shaft output of motor Ml is connected to the shaft output of motor M2. These two motors rotate in opposition and act as a differential drive for the vertical axis.

The body of motor M2 is connected to a bearing housing 30 carrying the horizontal axis of the robotic wrist. The horizontal axis of shaft 25 rotates synchronously with the body of motor M2. A bevel gear 28 is connected to the shaft 25 through the bearing housing 30. Differential rotational motion between motor M2 and motor Ml causes the bevel gear 28 to rotate and thus to cause the shaft carrying motor M3 to rotate. Motor M3 can then be made to rotate in opposition to the differential motion of the bevel gear 25 and so deliver a differential drive to the horizontal axis.

Thus the embodiment shown in Figure 5 provides a mechanism to generate a differential drive using 3 motors to provide two degrees of freedom for a Frenet Frame robotic wrist. Any competent practitioner can extend this to N + 1 motors driving N degrees of freedom.

Transparent simulation of flight from the UAVs perspective requires that the apparent velocity of the UAV with respect to the air is commensurate with the proper motion of the UAV experienced during flight. To achieve transparency there needs to be accurate control over the speed of the wind passing over the UAV under test. The dynamic control of wind speed is needed for transparency of the simulation. The generation of wind within a wind tunnel is an example of a transport delay. The speed of the wind is instantaneously determined by the dynamics of the fan. The speed of wind that passes the UAV under test is a delayed version of this speed due to the time taken for the air to move from the fan to the object under test.

Controlling systems with time delays is extremely difficult as the delay introduces a very large phase lag that prevents high gain control of the output of interest. High gain control is required of systems that are subject to uncertainty, particularly if there are disturbances within the transport phenomenon and the quality of the finished product is of importance. A known method used by control practitioners is to use a technique called a 'Smith Predictor' . Such a controller is not robust to uncertainty in a mathematical model and certainly will not be able to deliver sufficient control over wind tunnel transport delays to provide sufficient gain for transparency. Delays are a major limiter of performance. Figure 6 illustrates the principle of transport delay.

A transport delay is an infinite dimensional object, where the state is a description of the substance under transport at every instant of time. If it were possible to generate a finite dimensional description of the state of the substance being transported then alternative control methodologies become possible.

If the internal state of the transport phenomenon were to be measured at discrete points, then there is generated an equivalent discrete representation of the time delay. Uncertainties within the transport phenomenon can then be compensated for and much larger gains used on the control of the quality of the delivered substance. Sampling the flow at equidistant points effectively generates a delay line that encapsulates the internal state of the time delay as well as the effects of internal disturbances to the flow. The delay line generates a discretization of the time delay.

A general embodiment of how such a sampling system might be implemented is shown in Figure 7.

In the case of the dynamic testing of UAVs within a wind tunnel, the flow of interest is the flow of air. With reference now to Figure 2 the velocity of the air needs to be measured at multiple points between the source (the fan 15) and the UAV 1 under test so that a delay line can be implemented and the internal state of the air flow estimated.

An important issue regarding the achievement of transparency is the fine control over air speed and the support mechanism's interaction with the air flow. Actuators suffer from constraints regarding maximum rates of change. Linear control systems cannot easily deal with such actuators without significant detuning. In order to deal with constraints effectively, predictive control must be used. The delay line used to estimate the internal state of the transport time delay can be used to predict the future air flow over the UAV 1 and facilitates the control of both the support structure and the fan, where both are subject to actuator constraints. Fan control requires detailed information on the state of the system in order to achieve good prediction.

More particularly the signals from each pitot tube 16 instantaneously measure the velocity profile moving as one conceptually moves towards the fan 15. The signal from each tube 16 should accurately reflect the signal received at the next tube back down the flow towards UAV 1 after an appropriate delay. The real signals are sampled by the simulation control computer at a regular rate, say 500 times per second, so that the computer can calculate the expected flow rate for each tube in the future given the flow rate of the tube upstream. The consequent flow rate at the next sample time will not exactly match that calculated in the simulation control computer. Any such error adds information on the system internal state and thus can be used to update the computer's estimate of the actual state and well as how that state may be expected to evolve given the current state. In having such an accurate estimate of state and an accurate numerical model of how the state responds to inputs, changes to the speed of the fan can be calculated to achieve desired results much more accurately. Such a numerical model is called predictive. The computer can also adapt its behaviour to changes that it computes on how the speed of the fan influences the evolution of the velocity profile, such behaviour being called adaptive. If, in addition, there are probability distributions associated with parameters describing the model, and these distributions are identified by the computer during the operation of the system, then such a controller is called 'intelligent' or a 'learning' controller. None of these specialisations of control of the system are practically possible without robust estimation, which requires the delay line.

By way of illustration the arrangement of Figures 1 , 2 and 3 is used to test the landing autopilot of the UAV 1 on a moving aircraft carrier.

The memory associated with the onboard data processor 10 is provided with a landing autopilot program. The program is operative to cause the data processor 10 to control the servomotors (not illustrated) for the control surfaces in response to received radio signals as to the aircraft carrier's position and speed. The simulation control computer is programmed to interpret the control signals issued by the force-torque sensor 3 and convert those into control signals sent to the robotic wrist 2. The robotic wrist then implements the appropriate change of orientation to mimic that which the UAV 1 would have experienced during free flight. More particularly the force-torque sensor 3 measures the weight of the UAV 1 plus any forces imposed on it by the airstream. The aerodynamic forces are caused by the interaction of the control surfaces plus the general components of the airframe with the airflow. Such forces cause the UAV 1 to manoeuvre in free flight. If there is lift detected, for an accurate simulation the speed of the air must be reduced unless there is more power generated by the aircraft's prime mover. The controller must therefore compute the energy imparted to the air-stream, as well as the energy imparted to the aircraft, by virtue of the presumed velocity (stored as numbers within the simulation control computer) of the UAV 1 and the forces being imposed on it. This energy must be exactly matched in the simulation to prevent additional energy entering the simulation environment through poor transparency. The orientation of the wrist 2 must be such to impose the same expected Coriolis and centripetal forces that the UAV in free flight would experience - but to actually impose these as centrifugal forces via a D'Alambertian frame transformation. The simulation control computer 13 must therefore compute the expected kinetic energy of the vehicle, the expected loss due to drag and the expected increase in energy through changes in gravitational potential and input from the prime mover. These changes must then be made consistent with the orientation of the wrist 2 and speed of air and match the dynamic model of the UAV through the forces being sensed. These computations must be made at high speed and take into account uncertainty and time delays in the control of the air flow.

In the autopilot landing program scenario the simulation control computer 13 inputs data to the onboard data processor 5 which mimics GPS data of the moving aircraft carrier. It is to be noted that the illustrated connection between the simulation control computer 13 and the onboard data processor 5 by way of either a hard-wired connection or a wireless connection or a combination of both. Since the onboard data processor is provided with its initial location (in the simulated environment) and using its inertial navigation sensors 11 (stimulated by movements of the robotic wrist) and wind speed sensor 12 (stimulated by the flow of air generated by the fan 15) , the onboard data processor is able to calculate its current position. The data processor 5 is accordingly able to use the simulated GPS data and data pertaining to its own position and speed to calculate its speed and position in relation to the virtual aircraft carrier. The autopilot landing program is operative to control the course of the UAV 1 accordingly .

The actual speed of the wind at the UAV is measured using a Pitot sensor and is used to control the predictive component in the system. Linear acceleration must be simulated by causing the wind speed to vary in response to the attitude of the UAV, where turning is simulated by the roll and pitch of the three-axis wrist. The amount of roll and pitch will be determined by the curvature and torsion of the simulated path being followed by the UAV - the associated accelerations being replaced by D'Alembertian forces sensed in the force-torque sensor. The simulation control computer is configured to substantially synchronise a change in orientation with an appropriate change in wind speed as sensed by the UAV. Using a Kalman filter the computer 13 can accurately construct the state of the airflow with an accurate estimate of the airflow state upstream of the UAV and future airflow of the UAV can be predicted. The fan assembly 15 can be controlled accordingly.

Advantageously the inventive simulation apparatus can be used to test other automated flight programs, as well as the craft's ability to respond to pure or partial manual control. It is of particular importance to note that use of a plurality of sensors spaced in a downstream direction of a production source, in combination with a Kalman filter, finds utility in other areas in which transport delay is an issue. The sensors would measure an appropriate physical property at various positions along the direction of flow. For example use of a delay-line would find application in continuous feedstock delivery through a conduit in a process industry, extrusion through a die, rolling of steel where stock needs to pass through a plurality of rolling stages and the manufacture of thin films of plastic. The arrangement finds application in may other industrial processes.

Claims

1. Simulation apparatus for evaluation of a system's (1) performance, the apparatus comprising imposition means (2) which is adapted to be arranged in communication with the system (1) , simulation control means (13) and sensor means (3) , the apparatus being such that, in use, the system (1) is arranged to be in communication with the imposition means (2) and with the sensor means (3) , and the simulation control means (13) is operative to cause the imposition means (2) to act on the system (1) in response to a signal received from the sensor means (3) so as to provide a simulated effect corresponding to said signal.

2. Simulation apparatus as claimed in claim 1 which is an apparatus for evaluating a vehicle's (1) performance, the imposition means (2) comprising a motion simulator means, the apparatus being such that, in use, a vehicle is connected to the motion simulator and to the sensor means (3) and the sensor means is configured to sense activation of the motion control means (10) of the vehicle (1) and the simulation control means (13) is operative to cause the motion simulator means to act on the vehicle (1) to provide a simulated effect corresponding to the sensed activation of the vehicle's motion control means (10) .

3. Simulation apparatus as claimed in claim 2 in which the motion simulator means comprises a moveable support structure which is adapted to receive/support the vehicle, the simulation control means is configured to control the moveable support structure.

4. Simulation apparatus as claimed in claim 2 or in claim 3 which comprises controllable fluid flow means (15) , such that, in use, the vehicle is positioned in the path of the fluid flow.

5. Simulation apparatus as claimed in claim 2 in which the apparatus is suitable for evaluating an air vehicle, the apparatus comprising controllable fluid flow means (15), the motion simulator means comprising actuator means, the arrangement being such that, in use, an aerodynamic airframe is mounted on the motion simulator means, the motion simulator means is adapted to cause the airframe to adopt a series of desired orientations to simulate flight, and the airframe is positioned in the path of the fluid flow.

6. Simulation apparatus as claimed in claim 5 in which the sensing means (3) comprises force sensing means which is adapted to provide a measure of the interaction between the motion simulator means and the airframe.

7. Simulation apparatus as claimed in claim 6 in which on activation of a control surface of the airframe the simulation control means (13) , in use, is operative to convert a signal issued by the force sensing means to a control signal, the control signal causing the motion simulator means to exert on the airframe a movement substantially identical to that which would have been experienced during flight.

8. Simulation apparatus as claimed in claim 7 in which the force sensing means is disposed between the motion simulator means and the airframe .

9. Simulation apparatus as claimed in claim 6 in which the force sensing means comprises a six axis force-torque sensor.

10. Simulation apparatus as any of claims 6 to 9 in which, in use, the simulation control means is configured to use a signal emitted by the force sensing means to control the motion simulator to apply an appropriate force to the airframe.

11. Simulation apparatus as claimed in claim 2 in which the motion simulator means comprises a robotic support assembly having multiple degrees of rotational freedom.

12. Simulation apparatus as claimed in any of claims 4 to 11 further comprising a plurality of fluid flow sensors (16) which are provided in the flow path between the controllable fluid flow means (15) and the motion simulator means, each fluid flow sensor being configured to measure fluid flow rate.

13. Simulation apparatus as claimed in claim 12 in which fluid flow sensors (16) are spaced in the direction of fluid flow.

14. Simulation apparatus as claimed in claim 13 in which the fluid flow sensors (16) are substantially equi-distantly spaced.

15. Simulation apparatus as claimed in any of claims 12 to 14 in which the simulation control means (13) is configured to receive signals from the fluid flow rate sensors (16) and regulate the controllable fluid flow means (15) at least in part in response to said signals.

16. Simulation apparatus as claimed in claim 15 in which the simulation control means (13) is configured to process the signals received from the fluid flow rate sensors (16) using a Kalman filter.

17. Simulation apparatus as claimed in claim 16 in which the Kalman filter is used to estimate the flow fluid rate across the airframe.

18. Simulation apparatus as claimed in any preceding claim in which the simulation control means (13) regulates the controllable fluid flow means (15) at least in part in response to the instantaneous orientation of the airframe.

19. Simulation apparatus as claimed in claim 2 in which the motion simulator means comprises a differential drive arrangement in which two rotational drive means (Ml, M2; M3, M4) are provided for each rotational degree of freedom, and in use, the differential drive means associated with a degree of freedom are configured to rotate in opposite senses.

20. Simulation apparatus as claimed in claim 19 in which N rotational degrees of freedom are provided by N + 1 rotational drive means wherein at least one of the rotational drive means provides rotational input to at least two axes and N>2.

21. Simulation apparatus as claimed in claim 5 in which the simulation control means (13) is configured, in use, to issue simulated navigational data of a virtual target or destination to a control system (10) of the air vehicle.

22. Simulation apparatus as claimed in any preceding claim which is configured to evaluate a control system (10) of a vehicle.

23. Simulation apparatus as claimed in claim 22 in which the vehicle is an unmanned air vehicle (UAV) .

24. A method of evaluating a system's performance comprising arranging that the system (1) is in communication with imposition means (2) and with sensor means (3) , and causing the imposition means (2) to act on the system (1) in response to a signal received from the sensor means (3) so as to provide a simulated effect corresponding to said signal.

25. A method as claimed in claim 24 which is a method of evaluating an air vehicle, the method comprising locating an aerodynamic airframe on a motion simulator, causing the motion simulator to adopt a series of desired orientations so as to simulate flight and controlling the rate of fluid flow across the airframe

26. Apparatus for controlling a production process comprising a plurality of spaced-apart sensors (16) , the sensors being spaced in a downstream direction of a production source, the sensors being adapted to measure a physical aspect of a substance issued by the production source, the apparatus further comprising production source control means and the arrangement being such that, in use, the production source control means receives signals from the sensors and controls the production source at least in part in response to the received signals.

27. Apparatus for controlling a production process as claimed in claim 26 in which the production source control means comprises a data processor which is configured to implement a Kalman filter.

28. Apparatus for controlling a production process as claimed in either claim 26 or claim 27 in which the sensors comprise a plurality of fluid flow rate sensors.

29. Apparatus for controlling a production process as claimed in claim 28 in which the sensors comprise Pitot tubes.

30. Apparatus for controlling a production process as claimed in either of claim 26 or claim 27 in which the sensors are adapted to measure at least one of temperature, density, chemical concentration, physical dimensions, reflectance or electrical resistance.

31. A method of controlling a production process comprising receiving signals from a plurality of sensors, the sensors being spaced in a downstream direction of a production source, the signals being indicative of measures of a physical aspect of a substance issued by the production source, the method further comprising processing the signals and controlling the production source at least in part in response to the received signals.