Disclosure of Invention
In view of this, the invention provides an intelligent spacecraft controller based on the optical diffraction principle, which is composed of a plurality of modules including a signal receiving module, an electro-optical signal coding and converting module, a multilayer optical diffraction module, an electro-optical signal decoding and converting module and a signal sending module. The device can realize the interconversion and the encoding and decoding between an input and output electric signal and a nonlinear calculated optical signal, can flexibly and quickly realize the information bidirectional transmission between an optical diffraction calculation processing device and a traditional satellite borne processor, and can realize the functions of judgment, identification, classification, decision and the like of input parallel light carrying information based on the optical diffraction principle.
In order to achieve the purpose, the invention adopts the following technical scheme:
an intelligent spacecraft controller based on the principle of light diffraction, comprising: the system comprises a signal receiving module, an electro-optical signal coding and converting module, a multilayer optical diffraction module, an electro-optical signal decoding and converting module and a signal sending module which are connected in sequence; the electro-optical signal code conversion module receives the electric signal sent by the signal receiving module, codes the electric signal into an optical signal and transmits the optical signal to the multilayer optical diffraction module; the optical signal is transmitted to the photoelectric signal decoding and converting module through the multilayer optical diffraction module, and is converted into the electric signal through the photoelectric decoding and converting module, and the electric signal is transmitted to the signal sending module to be sent.
Preferably, the signal receiving module comprises a wired connection communication mode and a wireless connection communication mode; and the signal receiving module receives data information transmitted by an external processor. The signal receiving module receives satellite state information to be processed, capture point information, mechanical arm state information or lander landing point image information waiting processing data from the satellite-borne processor.
Preferably, the electro-optical signal coding module includes an electromagnetic wave generator, an electro-optical signal converter and an electronic information coding processing unit, the electromagnetic wave generator is connected to the electro-optical signal converter, the electro-optical signal converter is connected to the electronic information coding processing unit, the electronic information coding processing unit is connected to the signal receiving module to code the electrical signal and transmit the electrical signal to the electro-optical signal converter, the electrical signal is converted into the optical signal, and the optical signal is transmitted to the multilayer light diffraction module by the electromagnetic wave generator. The electromagnetic wave generator may be a coherent or incoherent electromagnetic wave generator; the electro-optical signal conversion device can adopt a spatial light modulator or a digital micro-reflector and the like; the photoelectric signal code conversion module is used for coding the data to be processed received by the signal receiving module so as to be suitable for light wave carrying coding, and carrying the coded data information to a carrier wave so as to be transmitted to the multilayer light diffraction module through an electromagnetic wave generator for optical calculation processing.
Preferably, the multilayer optical diffraction module comprises a plurality of layers of diffraction medium films which are arranged specifically, and the electromagnetic wave phase factor corresponding to each position of the diffraction medium films is determined; the diffraction medium film is mostly rectangular and thin, is made of a transparent material with uniform refractive index and can transmit electromagnetic waves such as laser and terahertz, and the plurality of layers of diffraction medium films can be fixed on the base in an equidistant or non-equidistant parallel arrangement mode; each layer of the diffraction medium film can be divided into a plurality of fine grids on a long and wide plane to form an optical diffraction network, the thickness of each grid on the diffraction medium film corresponds to different phase factors, the thickness of the diffraction medium film corresponding to each grid is obtained by solving the optimal value through a computer machine learning algorithm, the thickness of the diffraction medium film at different grids corresponds to different phase factors aiming at fixed-frequency light wave signals, and the effect that the thickness of different areas of the diffraction medium film is different can be realized through processes such as 3D printing or optical material etching. When the coherent parallel light beams penetrate through the diffraction medium films with different phase factors, the secondary spherical light waves generated at different grids of the diffraction medium films have different initial phases. When coherent parallel light passes through a plurality of layers of diffraction medium films fixedly arranged, a complex optical diffraction effect can be generated, and the process can realize an optical processing effect similar to that of a multilayer convolution neural network on the input coherent parallel light, namely, optical calculation is carried out on information of input light waves carried on electromagnetic waves such as laser or terahertz. The multi-layer diffraction medium film combination ensures that each layer of diffraction medium film has the same shape and size, the parallel relation is kept between the plates, and the central connecting line between the plates is vertical to the long and wide plane of any diffraction medium film, so that the light wave which penetrates through the optical diffraction network and is loaded with the calculation result can be projected to the photoelectric signal decoding and converting module. The long and wide planes are the main gathering planes overlapped by the plurality of layers of diffraction medium films.
Preferably, the photoelectric signal decoding and converting module comprises an imaging and capturing device and an electronic information decoding circuit; the imaging capture equipment captures the light waves carrying the calculation results, transmits the light waves to the electronic information decoding circuit to obtain the calculation results, decodes the calculation results to obtain real information, and transmits the real information to the signal sending module. The receiving plane of the imaging and capturing device is parallel to the plane (namely the length and width plane) with the largest area of the multilayer diffraction medium film, and the central line of the optical axis of the imaging and capturing device is collinear with the central connecting line of the multilayer diffraction medium film.
Preferably, the signal sending modules each include a wired connection communication mode and a wireless connection communication mode; and the signal sending module transmits the real information transmitted by the photoelectric signal decoding and converting module back to the external processor. And the signal sending module transmits the obtained calculation result corresponding to the satellite approaching capture track, the mechanical arm operation path or the lander landing track back to the satellite-borne processor so as to carry out perturbation operation on the corresponding satellite, the mechanical arm or the lander.
According to the technical scheme, compared with the prior art, the invention discloses and provides the spacecraft intelligent controller based on the light diffraction principle, the spacecraft intelligent controller realizes the functions similar to those of the traditional nonlinear planner on the basis of the light diffraction network in the multilayer light diffraction module, can encode input state information and carry the encoded input state information on the light wave, and the light wave carrying the information can decode the information of an output platform after passing through the multilayer optical diffraction network to obtain the optimal planned path with the result similar to that of the traditional nonlinear planner. The main working scenes of the invention are scenes which need real-time path planning, such as space satellite approach accompanying flight, mechanical arm grabbing, detector landing and the like, and path planning operation including nonlinear planning can be carried out in real time at high speed and low power consumption. The controller encodes information such as the flight state of the spacecraft, the state of a mechanical arm and a target, the landing state of the lander and the like and carries the encoded information on the parallel light, the parallel light carrying the information passes through the multilayer optical diffraction network, and then the information of the light of the output plane is decoded to obtain relevant path prediction information required by functions such as docking, accompanying, capturing, landing and the like of the spacecraft.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The embodiment of the invention discloses a spacecraft intelligent controller based on a light diffraction principle, which comprises: the system comprises a signal receiving module, an electro-optical signal coding and converting module, a multilayer optical diffraction module, an electro-optical signal decoding and converting module and a signal sending module which are connected in sequence; the electro-optical signal code conversion module receives the electric signal sent by the signal receiving module, codes and converts the electric signal into an optical signal and transmits the optical signal to the multilayer optical diffraction module; the optical signal is transmitted to the photoelectric signal decoding and converting module through the multilayer optical diffraction module, is converted into an electric signal through the photoelectric decoding and converting module, is transmitted to the signal sending module and is sent to the signal sending module.
In order to further optimize the technical scheme, the signal receiving module comprises a wired connection communication mode and a wireless connection communication mode; the signal receiving module receives data information transmitted by the external processor. The signal receiving module receives satellite state information to be processed, capture point information, mechanical arm state information or lander landing point image information waiting processing data from the satellite-borne processor.
In order to further optimize the technical scheme, the electro-optical signal coding module comprises an electromagnetic wave generator, an electro-optical signal converter and an electronic information coding processing unit, wherein the electromagnetic wave generator is connected with the electro-optical signal converter, the electro-optical signal converter is connected with the electronic information coding processing unit, the electronic information coding processing unit is connected with the signal receiving module to code electric signals and transmit the electric signals to the electro-optical signal converter, the electric signals are converted into optical signals, and the optical signals are transmitted to the multilayer light diffraction module through the electromagnetic wave generator. The electromagnetic wave generator may be a coherent or incoherent electromagnetic wave generator; the electro-optical signal conversion device can adopt a spatial light modulator or a digital micro-reflector and the like; the photoelectric signal coding and converting module is used for coding the data to be processed received by the signal receiving module so as to be suitable for light wave carrying coding, and carrying the coded data information to a carrier wave so as to be transmitted to the multilayer light diffraction module through an electromagnetic wave generator to carry out optical calculation processing.
In order to further optimize the technical scheme, the multilayer optical diffraction module comprises a plurality of layers of diffraction medium films which are arranged specifically, and the electromagnetic wave phase factor corresponding to each position of each diffraction medium film is determined; the diffraction medium film is mostly rectangular in shape and thin in thickness, the material is transparent and is made of materials with uniform refractive index, the material can transmit electromagnetic waves such as laser and terahertz, and the multilayer diffraction medium film can be fixed on the base in an equidistant or non-equidistant parallel arrangement mode. Each layer of diffraction medium film can be divided into a plurality of fine grids on the long and wide plane to form an optical diffraction network, different thicknesses of the diffraction medium film correspond to different phase factors, the thickness of the diffraction medium film corresponding to each grid is optimally obtained by solving a computer machine learning algorithm, the thicknesses of the diffraction medium films at different grids correspond to different phase factors aiming at fixed-frequency light wave signals, and different thicknesses of different areas of the diffraction medium film can be manufactured through processes such as 3D printing or optical material etching. When the coherent parallel light beams penetrate through different diffraction medium films at different phase factors, the secondary spherical light waves generated at different grids of the diffraction medium films have different initial phases. When coherent parallel light passes through a plurality of layers of diffraction medium films which are fixedly arranged, a complex optical diffraction effect can be generated, and the process can realize an optical processing effect similar to that of a multilayer convolution neural network on the input coherent parallel light, namely, optical calculation is carried out on information of input light waves carried on electromagnetic waves such as laser or terahertz. The multilayer diffraction medium film combination ensures that each layer of diffraction medium film has the same shape and size, the parallel relation is kept between the plates, the central connecting line between the plates is kept vertical to the long and wide plane of any diffraction medium film, and further ensures that light waves carrying calculation results and transmitted through the optical diffraction network can be projected to the photoelectric signal decoding and converting module. The long and wide planes are the main collecting surfaces of the overlapped parts of the multilayer diffraction medium films.
In order to further optimize the technical scheme, the photoelectric signal decoding and converting module comprises an imaging capture device and an electronic information decoding circuit; the imaging capture equipment captures the light waves carrying the calculation results, transmits the light waves to the electronic information decoding circuit to obtain the calculation results, decodes the calculation results to obtain real information and transmits the real information to the signal sending module.
In order to further optimize the technical scheme, the signal sending modules comprise a wired connection communication mode and a wireless connection communication mode; and the signal sending module transmits the real information transmitted by the photoelectric signal decoding and converting module back to the external processor. And the signal sending module transmits the obtained calculation result corresponding to the satellite approaching capture track, the mechanical arm operation path or the lander landing track back to the satellite-borne processor so as to carry out perturbation operation on the corresponding satellite, mechanical arm or lander.
Example 1
Approaching of the failed satellite and catching of the mechanical arm:
when the satellite on the circular orbit or the elliptical orbit fails, the ground needs to transmit the service satellite to be close to the corresponding failed satellite and operate the mechanical arm on the service satellite to capture the failed satellite.
The flight history for a failed satellite approach acquisition can be roughly divided into 4 phases. Phase 1, the orbital transfer phase, the service satellite is first manoeuvred to a suitable position (100km) by the mother satellite carrying it, then separated from the mother satellite, the orbital transfer is manoeuvred to the vicinity of the target (for example within a hundred meters), and then the approach procedure is started. And 2, finally, in the approach section, starting the laser radar and the optical camera, acquiring the position and posture information of the target and the appropriate capture point on the target, continuously keeping the approach, and judging whether to implement capture. And 3, in the capturing section, unlocking the mechanical arm at a proper time, keeping the service satellite close until the service satellite enters the relatively static flying-around state (or flying-by state), and simultaneously extending the mechanical arm to complete contact capturing of a capturing point. And 4, in a leaving section, if the capturing cannot be successfully carried out or is in danger (such as the impact is too large or the mechanical arm is possibly pulled off), the service satellite does not unlock the mechanical arm and controls the service satellite to leap by the target.
Wherein the light diffraction controller of the present invention takes part in the above 2 nd and 3 rd stages.
In stage 2:
s11: the laser radar and the optical camera are started to obtain the position and posture information of the failed satellite and the proper capture point of the failed satellite, and the satellite-borne processor sends the corresponding posture information to the light diffraction controller;
s12: the light diffraction controller realizes that: encoding pose information and carrying the pose information to input light waves, decoding the input light waves through a multilayer light diffraction network and a decoder to output plane light wave information to obtain optimal planning path information of a corresponding service satellite and sending the optimal planning path information to a satellite-borne processor;
s13: and the satellite borne processor carries out satellite body perturbation operation according to the optimal planning path information.
In stage 3:
s21: after the mechanical arm is unlocked, the laser radar and the optical camera acquire position and posture information of a position where the failed satellite is properly captured, and the satellite-borne processor sends state information of the mechanical arm and the position and posture information of the capture position to the light diffraction controller;
s22: the light diffraction controller realizes that: the state information and the pose information of two satellites of the mechanical arm are coded and carried to an input light wave, the input light wave is decoded by a multilayer light diffraction network and a decoder to output plane light wave information, optimal planning path information of the corresponding mechanical arm is obtained and sent to a satellite-borne processor;
s23: and the satellite-borne processor controls the mechanical arm to carry out optimal execution operation according to the optimal planning path information of the mechanical arm.
Example 2
Calculating the optimal landing track of the planet detector:
in the landing link of the lunar and mars detectors, the star surface environment (terrain, geology and the like) cannot be comprehensively detected, the initial landing place may not be the optimal landing point, and the landing track of the detector is detected and finely adjusted in real time in the landing process of the detector. The specific experimental scheme is as follows:
s31: a star catalogue probe (lander) starts landing operation according to an initial preset landing point target;
s32: a detector (lander) carries a laser radar and a visible light camera to capture a landing point image corresponding to a current landing track and sends integrated data to a light diffraction controller;
s33: the light diffraction controller realizes that: encoding the image information of the expected landing point of the current path and carrying the image information to input light waves, carrying out light diffraction processing on the input light waves through a light diffraction network of a multilayer light diffraction module, decoding and outputting plane light wave information by a decoder of a photoelectric signal decoding and converting module to obtain the optimal landing track of the corresponding lander and sending the optimal landing track to an on-board processor;
s34: the spaceborne processor determines the perturbation direction and the corresponding perturbation quantity of an engine of the lander according to the optimal landing track of the lander calculated in real time;
s35: and iterating the processes of S32-S34 for multiple times until the lander completes the star landing task.
The above embodiments 1 and 2 are different in point of importance in that the input information is carried on the optical wave. The input information of the embodiment 1 is pose information in a satellite perturbation or mechanical arm operation process, the pose information is obtained by calculation in a satellite processor according to images captured by a laser radar or a visible light camera, so that the input of a light diffraction controller is a series of pose information, and the light diffraction controller realizes a fitting function of a nonlinear complex function; in embodiment 2, the input information of the optical diffraction controller is a composite image of a landing point of a predetermined trajectory of the lander, and the output information is an updated optimal planned trajectory, so that the optical diffraction controller implements a saliency detection function based on the image.
Example 3
The multilayer light diffraction module of the spacecraft intelligent controller comprises a plurality of layers of diffraction medium films which are specifically arranged, the electromagnetic wave phase factor corresponding to each position of each diffraction medium film is determined, the electromagnetic wave phase factor of each diffraction medium film is determined by the thickness of the diffraction medium film, and the thickness calculation method of each diffraction medium film in the multilayer diffraction medium film structure in different areas comprises the following specific processes:
as shown in fig. 2, according to the fourier optical theory and the huygens principle, after the input light wave passes through the i-th diffractive dielectric film, different pixel regions on the i-th diffractive dielectric film can be regarded as a secondary spherical wave source, and compared with phase parameters of the input light wave in different pixel regions of the i-th diffractive dielectric film, phase parameters of a new spherical wave source at a corresponding pixel region on the i-th diffractive dielectric film are modulated by a phase factor of the corresponding pixel region, and a certain variable exists. The secondary spherical waves of different pixel areas after the ith diffraction medium film are diffracted on the (i + 1) th diffraction medium film, so that the light wave parameters of the previous diffraction medium film received by the specific pixel area on the (i + 1) th diffraction medium film are obtained by superposing the light vectors of all the secondary spherical waves of the ith layer at the position. That is to say, like a conventional convolutional neural network, the light wave signals received by the pixel region of the (i + 1) th layer of diffraction medium film are obtained by superposing all secondary spherical waves of the i-th layer of diffraction medium film, so that mathematical formula description of propagation of the light wave signals among each layer of diffraction medium film can be established, phase parameters of a generated secondary wave source after input light waves pass through each layer of diffraction medium film are obtained by calculation in sequence, and light intensity distribution of spherical waves generated by the secondary wave source of the last layer of diffraction medium film on an output plane is obtained.
The model training method for solving the corresponding thicknesses of different grids of the diffraction medium film based on computer machine learning comprises the following steps:
s1: according to the input signal, designing the light intensity distribution of the output light wave carrying the output signal on an output plane, namely a true value;
s2: initializing each layer of light diffraction medium film thickness parameter (namely corresponding phase factor), and making the input light wave pass through each layer of diffraction medium film to obtain the light intensity distribution of the output plane under the current network parameter;
s3: designing a loss function, and calculating the difference between the light intensity distribution of the current output plane and the light intensity distribution of the true value;
s4: the method is suitable for updating parameters such as back propagation, chain type derivation and the like, and iteratively updates the thickness parameter of each layer of diffraction medium film to realize one-time training;
s5: selecting a next batch of training samples, and training and optimizing network parameters;
s6: and repeating S1-S5 until the loss function of the optical diffraction network does not continuously decrease, or the number of network iterations reaches the requirement of a set value, or the loss function meets the requirement of the set value, and finishing the training.
Through the steps 1-6, the training calculation process of different area thicknesses (phase factors) of each layer of diffraction medium film of light diffraction can be realized aiming at a specific application scene and a training test sample set, namely the training process of the light diffraction network is realized.
The photoelectric signal code conversion module of the spacecraft intelligent controller carries out coding processing on data to be processed received by the signal receiving module so as to be suitable for light wave carrying codes, carries coded data information to a carrier wave so as to be transmitted to the multilayer light diffraction module through the electromagnetic wave generator to carry out optical calculation processing, and the multilayer diffraction module carries out optical calculation on the information carried on input light waves in the following calculation process:
just as according to the fourier optical theory and the huygens principle, by setting the thickness parameters (phase factors) of different pixel regions of each layer of diffraction medium film, mathematical formula description of propagation of the light wave signal between each layer of diffraction medium film can be established, phase parameters of the generated secondary wave source after the input light wave passes through each layer of diffraction medium film are sequentially calculated, and the light intensity distribution of the spherical wave generated by the secondary wave source of the last layer of diffraction medium film on the output plane is obtained. The optical diffraction network is similar to a traditional convolution neural network and has a function of fitting a function, and compared with the step of receiving the light intensity distribution by the output plane, nonlinearity is added to the network, so that the network has the performance of fitting a nonlinear complex function.
Therefore, the input quantity of the optical diffraction network controller can be encoded on the phase of the input light wave and the spatial position distribution of the light waves with different phases by a specific encoding mode; and through a specific coding mode, the expected output quantity of the controller is coded on the light intensity distribution of the output plane, and the training of optical network parameters is carried out, namely, the decoding process from the light intensity distribution of the output plane to the output quantity of the controller can be realized. Because the input light wave and the expected output plane optical distribution of the optical diffraction network are preset and are represented optically from the input and output variables of the traditional nonlinear controller, the training process of the multilayer optical diffraction network is the fitting approximation process of the traditional controller nonlinear controller. After the input light wave carrying the input information of the controller passes through the multilayer optical diffraction network, a specific light intensity distribution is obtained on an output plane, and the optical diffraction controller is made to fit the input-output relation of the traditional nonlinear controller to the maximum extent through training, so that the light intensity distribution of the light wave passing through the optical diffraction network on the output plane is considered to be the optical representation output by the fitted nonlinear controller.
The invention integrates the path planning module which originally needs the large amount of calculation and power consumption of the on-board processor into the optical diffraction controller module, shares the calculation burden of the on-board processor on the basis of ensuring that the model and the performance of the original on-board processor are unchanged, reduces the power consumption loss in the path planning process of the spacecraft, and improves the real-time performance of the path planning of the spacecraft.
According to the spacecraft intelligent controller based on light diffraction, a core component is a multilayer optical diffraction medium film, phase factors (thickness of a region corresponding to the medium film) corresponding to different positions of each layer of medium film are obtained through repeated iterative training of a ground computer, and the phase factors are fixed and do not change after the medium film is produced, so that the optical diffraction medium film component in the intelligent processor is not easily influenced by electromagnetic electronic interference such as outer space radiation and single event effect, and has stronger anti-interference capability compared with a large number of electronic circuits in a traditional satellite-borne processor.
According to the Fourier optical theory, when parallel light passes through different dielectric films at phase factors, phase changes of different degrees, namely diffraction of different degrees, can occur; when parallel light passes through a plurality of dielectric films with different phase factors, the process can achieve the optical processing effect similar to that of a multilayer convolution neural network on input parallel light. By training and determining phase factor values at different positions of each layer of dielectric film, an intelligent controller based on an optical diffraction network can be obtained, and functions of classification, fitting, decision making and the like can be realized on information carried by input parallel light. The information such as the flight state of the spacecraft, the state of a mechanical arm and a target, the landing state of the lander and the like can be coded and carried on the parallel light, the parallel light carrying the information passes through the multilayer optical diffraction network controller, and the related path prediction information required by functions such as docking, accompanying, capturing, landing and the like of the spacecraft can be obtained by decoding the light information of the output plane. Compared with the traditional electronic-based spaceborne processor, the optical diffraction spaceborne controller has the characteristics of low power consumption, high computing power, real-time performance, interference resistance and the like.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.