Artificial intelligence

Abstract

Autonomous control of an entity such as a UAV or unmanned aircraft is used to perform an action. Control is based upon received situational awareness data and a generated random number. A control sign

Description

FIELD The present disclosure relates to an apparatus for autonomous control of an entity. The present disclosure also relates to a method of autonomously controlling an entity. BACKGROUND Vehicles and systems controlled to some degree by artificial intelligence are known. Such vehicles include, for example, unmanned aircraft. In conflict scenarios these vehicles tend to be at a disadvantage as adversaries may be able to predict their actions and reactions. Therefore, there is a need to provide an improved decision-making tool and method of controlling a vehicle or other entity that addresses at least this problem. SUMMARY According to a first aspect of the present disclosure, there is provided an apparatus for controlling an entity to perform an action, the apparatus comprising: an interface for receiving situational awareness data; and a controller configured to: receive a random number; and generate a control signal to control the entity to perform an action based on the situational awareness data and the random number. Advantageously, by using random numbers to influence actions to be performed, an entity tends to be unpredictable. This is particularly advantageous in combat situations or those where an entity, such as a vehicle, faces an opponent. The controller may be configured to generate a list of possible actions based on the situational awareness data, and use the random number to select one of the actions to perform. The apparatus may comprise a random number generator for generating the random number. Alternatively, the random number may be received through an interface from a second entity. The random number generator may comprise a Physically Unclonable Function [PUF], Alternatively, the random number generator may comprise a radioactive source and a radiation detector, wherein the radiation detector is configured to measure a time between a plurality of detected emissions and generate a random number based on the plurality of times. Alternatively again, the random number generator may comprise an internal clock used for generating a random number. The apparatus may comprise means for reading the PUF and generating data representative of the PUF, wherein the random number generator is configured to generate a random number based on the data. The means for reading the PUF may comprise an image capture device. The PUF may be a semiconductor. The entity may comprise a vehicle. The action may comprise a manoeuvre or other tactical decision. For example, the action may comprise the release of a payload. Alternatively, the action may comprise changing the speed, altitude or direction of the entity. Alternatively again, the action may comprise changing a communication characteristic of the entity, such as the antenna gain. According to a second aspect of the present invention, there is provided a vehicle comprising the apparatus according to the first aspect. The vehicle may be an unmanned aircraft. According to a third aspect of the present invention, there is provided a method of controlling an entity comprising: receiving situational awareness data; receiving a random number; and generating a control signal to control the entity to perform an action based on the situational awareness data and the random number. The method may comprise generating a list of possible actions based on the situational awareness data, and using the random number to select one of the actions to perform. The method may comprise generating the random number. Alternatively, receiving the random number may comprise receiving the random number from a second entity. The method may comprise measuring the time between radioactive emissions and generating a number based on a plurality of measured times. The method may comprise reading a PUF, generating data representative of the PUF, and generating a random number based on the data. Reading the PUF may comprise capturing an image of the PUF. The action may be the release of a payload. Alternatively, the action may comprise manoeuvring a vehicle. Alternatively again, the action may comprise changing the speed, altitude or direction of the entity. Alternatively again, the action may comprise changing a communication characteristic of the entity, such as the antenna gain. It will be appreciated that features described in relation to one aspect of the present disclosure can be incorporated into other aspects of the present disclosure. For example, an apparatus of the disclosure can incorporate any of the features described in this disclosure with reference to a method, and vice versa. Moreover, additional embodiments and aspects will be apparent from the following description, drawings, and claims. As can be appreciated from the foregoing and following description, each and every feature described herein, and each and every combination of two or more of such features, and each and every combination of one or more values defining a range, are included within the present disclosure provided that the features included in such a combination are not mutually inconsistent. In addition, any feature or combination of features or any value(s) defining a range may be specifically excluded from any embodiment of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS Embodiments of the disclosure will now be described by way of example only and with reference to the accompanying drawings. Figure 1 is a system diagram of an unmanned vehicle according to an embodiment of the present disclosure; Figure 2 is a system diagram of an unmanned vehicle according to another embodiment of the present disclosure; and Figure 3 is a flowchart showing a method of controlling a vehicle according to an embodiment of the present disclosure. For convenience and economy, the same reference numerals are used in different figures to label identical or similar elements. DETAILED DESCRIPTION Generally, embodiments herein relate to vehicles that rely on artificial intelligence (Al) to perform functions. The embodiments are particularly suitable for military vehicles, such as unmanned combat air vehicles, which operate in an environment where predictability can put a vehicle at a disadvantage. For example, through espionage or previous encounters with a typical vehicle controlled by Al, an adversary may be able to determine what the vehicle will do in any given situation and therefore be prepared to counter or pre-empt the vehicle’s actions. The solution presented in the embodiments that follow is to introduce an unpredictability to the Al through a random number generator. In order to make the number truly random, and hence the Al entirely unpredictable, a Physical Unclonable Function (PUF) is used. It would be counterintuitive to modify every function of a vehicle such that they perform in an unpredictable manner. For example, civilian driverless cars must have rules that define exactly what the car will do as it approaches a red traffic light, or identifies an object in the road, in order to conform to legislation or provide a degree of safety. However, prior art vehicles having a degree of autonomy have been designed in a prejudiced manner to rely entirely on this predictability, which is not always beneficial. Neural networks are a type of Al whereby a controller is “trained” how to behave in certain situations, but over time the actions performed by the controller become predictable when faced with the same problem repeatedly. Figure 1 shows a system diagram of a vehicle 10 according to embodiments. The vehicle 10 is an unmanned aircraft according to a preferred embodiment. However, it would be appreciated that in other embodiments the vehicle 10 may be a land vehicle, surface ship or submarine. The vehicle 10 may be manned, unmanned or optionally-manned. The vehicle 10 includes a controller 12. The controller 12 may take any suitable form, such as a microcontroller, plural microcontrollers, processor or plural processors. The controller 12 is used to control a number of on-board systems autonomously using artificial intelligence. The systems controlled by the controller 12 include but are not limited to a communication system 24, a propulsion system 25, and a flight control system 26. Further systems controllable by the controller 12 may include weapons systems. The controller 12 may be a software-based decision-making module for selecting an action to be performed by a controller in which it is embedded, or a separate controller. The communication system 24 may be controlled by the controller 12 to transmit a request to other vehicles for assistance. Alternatively, the communication system 24 may be controlled to reduce power to an antenna, or turn off entirely, in response to a threat. Each of these optional actions may be assigned a number, and a generated random number then used to selection an action to perform. Similarly, the controller 12 may be used to select from a range of weapons systems. For example, the controller 12 may select to engage a target with guns or with short-range air-to-air missiles, depending on a generated random number. While a certain level of predictability, or even human input, is desired when deciding whether or not to engage a target, how that target is to be engaged could be made unpredictable. For example, if an adversary always knows that missiles will be deployed if it is within a certain range of the vehicle 10, then it can prepare itself with countermeasures or manoeuvres. The controller 12 may control the vehicle 10 to speed up by adjusting the propulsion system 25 in order to escape a scenario, or may control the vehicle 10 to slow down to engage a target, depending on a generated random number. Similarly, the controller 12 may control the vehicle’s heading or altitude to evade or engage a threat depending on a generated random number, using the flight control system 26. The controller 12 receives situational awareness information, such as the location or type of threat, or the presence of clouds, from at least one sensor 16. The sensor 16 may be an optical camera, for example. The sensor 16 may also be a radar receiver. The controller 12 may be configured to control the sensor 16 to change pointing direction or focal length, or increase or decrease the power of transmission if it is an active device (as opposed to passive). The sensor 16 may be another type of receiver, such as an RF receiver, infrared receiver, or hyperspectral imager. The controller 12 receives information about the vehicle 10 from a navigation system 14. The navigation system 14 may include a satellite receiver, such as GPS or GLONASS. The navigation system 14 may comprise a gyroscope and / or compass for determining the heading or pitch of the vehicle 10. The navigation system 14 may comprise an airspeed indicator. The controller 12 is coupled to a memory 18 (or electronic storage medium). The memory 18 may be a non-volatile memory such as read only memory (ROM) a hard disk drive (HDD) or a solid state drive (SSD). The memory 18 stores code, which when executed by the controller 12, controls operation of each of hardware components of the vehicle 10. The memory 18 may further comprise volatile memory such as random access memory (RAM), used by the controller 12 for the temporary storage of data. In one embodiment, as illustrated in Figure 1, the memory 18 includes an algorithmic random number generator 17 for execution by the controller 12. In an example of an algorithmic random number generator, the time in milliseconds is used as a seed for a function to produce random numbers. The present embodiment is advantageous in that it generally provides the vehicle 10 with unpredictability. However, it has the disadvantage that as soon as a malicious actor identifies that a time-based algorithm is being used to generate random numbers that dictate how the vehicle 10 will operate, all they have to do is determine the time-dependent function being used by the algorithm to can gain access to anything encrypted with the random numbers or predict vehicle 10 responses. The function may be readily determined through quantum computing. In another embodiment, illustrated in Figure 2, the vehicle comprises a Physically Unclonable Function (PUF) 22. The PUF 22 may, for example, be an imperfect surface. An imperfect surface is one that has naturally-occurring non-replicable randomness, for example at the submicroscopic scale. Examples of a PUF 22 include quantum PUFs (QPUFs), SRAM PUFs, controlled PUFs (CPUFs), radio frequency PUFs (RF-PUFs), and optical-PUFs (e.g. a semiconductor material), which would be appreciated by the skilled person. PUFs 22 may be biological in nature. There are a number of different types of PUFs 22 that could be used on the vehicle 10, which would be known to the skilled person. An optical PUF 22, as illustrated in Figure 2, tends to result in reduced physical connection to the controller 12 hardware and therefore increased cyber security. Further, in some embodiments, the vehicle 10 includes a PUF reader 20 for reading the PUF 22. The PUF reader 20 generates data for use by the controller 12 in generating a random number. The PUF reader 20 may be, for example a laser beam that is used to scan the PUF 22 and record microscopic undulations in its surface. The PUF reader 20 may be an optical camera for identifying changes in patterns or colours in the PUF 22. The camera may be coupled to a microscope, for example an electron microscope, such that variations in microscopic structure of the PUF 22 can be identified. In another embodiment, the PUF 22 is integrated within circuitry of the controller 22 (or other random number generating unit). Here, variations in electrical properties of the circuitry are read by the PUF reader 20. The PUF reader 20 comprises electronic components such as ammeters or voltmeters in order to generate data for generating a random number. In another embodiment, the physical PUF 22 is read by a device that is not part of the vehicle, and data representative of the PUF 22 is stored in memory 18. The image of the PUF 22, or data representing the PUF 22, is stored in the memory 18 such that the controller 12 can use it to generate a random number. Optionally, the data representing the PUF 22 may be divided to give the effect of a plurality of PUFs 22 in the memory 18. In each embodiment, the random number acts as a seed used by the controller 12 to select decisions from a list of options. The controller 12 may generate a plurality of options for controlling systems 24, 25, 26 of the vehicle 10 in a given tactical situation. The random number generated by the random number generator is then used to select one of those options. The list of options may be stored in the memory 18. By using a truly random number, there is no chance of repeatability or predictability. Using a PUF 22 as a random number generator tends to provide two benefits. Firstly, the number generated will be a truly random and unique number. Secondly, PUFs 22 by their intrinsic nature tend to be extremely resilient to being hacked, copied and / or duplicated. The PUF 22 may be divided into multiple tags, each bearing its own individual random numbers. Different tags could be read by the PUF reader 20 at different points in a mission (for instance, the PUF reader 20 may read a new tag after a predetermined time has passed) to further increase the randomness of decisions made. Each platform would have its own set of PUF tags resulting in a high probability of zero repeatability of patterns in decision-making. Another alternative embodiment provides a radioactive source in place of the PUF 22. The radioactive source is then used by the controller 12 to generate a random number. Radioactive emission, such as of alpha particles, beta particles or neutrons, tends to be unpredictable, and therefore by recording the time between detections a random number can be generated. This embodiment is not optimal when compared with the use of a PUF 22, as the radioactive source and associated containment and measurement equipment tends to be physically significantly larger, more expensive and has more safety concerns. Further, it is also a slower method of retrieving a random number. A method of controlling the vehicle 10 will now be described with reference to Figure 3. In a first step, S300, the controller 12 receives situational awareness data from the at least one sensor 16. This may include, for example, the type, heading and location of a threat (i.e. adversary unit or vehicle). In step S302, the controller 12 determines a list of potential actions that could be performed in response to the present situation. These options may include, for example, changing course, altitude or speed of the vehicle 10, or deploying different types of weapon. At step S304, the decision-making module of the controller 12 receives a random number. The random number may be received by the controller 12 from another module within the controller 12, an independent number generation means within the vehicle 10, or the random number may be received from elsewhere. In one embodiment, the random number may be generated by the controller 12 in step S303 by reading (e.g. scanning, measuring, photographing) a PUF 22 using a PUF reader 20 such as an imaging device. Alternatively, the random number may be generated in step S303 by an algorithm, such as a time-based algorithm. The random number may be prestored in a memory 18. Alternatively, the random number may be received from a ground station or other vehicle, via an interface. At step S306, the controller 12 uses the received random number as a seed to select one of the options identified in step S302. This may comprise providing each option with a unique number, converting the random number into an integer value within the range of unique numbers, and selecting the option having the unique number closest to the integer value. In other words, the controller 12 performs the function of an unpredictable artificial intelligence decision-making tool. In step S308, the controller 12 controls one or more vehicle systems 24, 25, 26 to perform the option selected in step S306. While the present disclosure has been described in relation to a vehicle 10, the skilled person would appreciate that it is equally applicable to other entities. For example, in another embodiment, the controller 12 is incorporated into a desktop computer. Here, the controller 12 is used to provide unpredictability to a games engine such that Al-controlled characters or actors in a game do not perform the same action every time under the same or similar circumstances. Here, situational awareness data is itself artificial, and includes for example the distance to the next friendly player in a football game or the distance of and direction to a threat in a combat game. Where, in the foregoing description, integers or elements are mentioned that have known, obvious, or foreseeable equivalents, then such equivalents are herein incorporated as if individually set forth. Reference should be made to the claims for determining the true scope of the present disclosure, which should be 5 construed so as to encompass any such equivalents. It will also be appreciated by the reader that integers or features of the disclosure that are described as optional do not limit the scope of the independent claims. Moreover, it is to be understood that such optional integers or features, while of possible benefit in some embodiments of the disclosure, may not be desirable, and can therefore 10 be absent, in other embodiments.

Claims

1. An apparatus for autonomously controlling an entity to perform an action, the apparatus comprising:an interface for receiving situational awareness data; anda controller configured to:receive a random number; andgenerate a control signal to control the entity to perform an action based on the situational awareness data and the random number, and,wherein the action comprises a tactical decision comprising at least one of: the release of a payload; and, changing a communication characteristic of the entity.

2. The apparatus according to claim 1, wherein the controller is configured to generate a list of possible actions based on the situational awareness data, and use the random number to select one of the actions to perform.

3. The apparatus according to claim 1 or claim 2, comprising a random number generator for generating the random number.

4. The apparatus according to claim 3, wherein the random number generator comprises a nuclear source and a radiation detector, wherein the radiation detector is configured to measure a time between a plurality of detected emissions and generate a random number based on the plurality of times.

5. The apparatus according to claim 3, wherein the random number generator comprises a Physically Unclonable Function [PUF],6. The apparatus according to claim 5, comprising a means for reading the PUF and generating data representative of the PUF, wherein the random number generator is configured to generate a random number based on the data.

7. The apparatus according to claim 6, wherein the means for reading the PUF comprises an image capture device.

8. The apparatus according to any one of claims 5 to 7, wherein the PUF is a semiconductor.

9. The apparatus according to any one of the preceding claims, wherein the entity comprises a vehicle.

10. A vehicle comprising the apparatus according to any one of the preceding claims.

11. The vehicle according to claim 10, wherein the vehicle is an unmanned aircraft.

12. A method of autonomously controlling an entity, comprising:receiving situational awareness data;receiving a random number; andgenerating a control signal to control the entity to perform an action based on the situational awareness data and the random number, and,wherein the action comprises a tactical decision comprising at least one of: the release of a payload; and, changing a communication characteristic of the entity.

13. The method according to claim 12, comprising generating a list of possible actions based on the situational awareness data, and using the random number to select one of the actions to perform.

14. The method according to claim 12 or claim 13, comprising generating the random number.

15. The method according to claim 14, comprising measuring the time between radioactive emissions and generating a number based on a plurality of measured times.5 16. The method according to claim 14, comprising reading a PUF, generatingdata representative of the PUF, and generating a random number based on the data.

17. The method according to claim 16, wherein reading a PUF comprises 10 capturing an image of the PUF.CMCM 15

Images