CN101733746A - Autonomously identifying and capturing method of non-cooperative target of space robot - Google Patents

Abstract

The invention relates to an autonomously identifying and capturing method of a non-cooperative target of a space robot, comprising the main steps of (1) pose measurement based on stereoscopic vision, (2) autonomous path planning of the target capture of the space robot and (3) coordinative control of a space robot system, and the like. The pose measurement based on the stereoscopic vision is realized by processing images of a left camera and a right camera in real time, and computing the pose of a non-cooperative target star relative to a base and a tail end, wherein the processing comprises smoothing filtering, edge detection, linear extraction, and the like. The autonomous path planning of the target capture of the space robot comprises is realized by planning the motion tracks of joints in real time according to the pose measurement results. The coordinative control of the space robot system is realized by coordinately controlling mechanical arms and the base to realize the optimal control property of the whole system. In the autonomously identifying and capturing method, a self part of a spacecraft is directly used as an identifying and capturing object without installing a marker or a comer reflector on the target star or knowing the geometric dimension of the object, and the planned path can effectively avoid the singular point of dynamics and kinematics.

Description

Non-cooperative target of space robot autonomous classification and catching method

Technical field

The present invention relates to a kind of non-cooperative target of space robot autonomous classification and catching method, belong to robot for space in rail service technology field.

Background technology

Along with development of technology, activity of human beings is in space-ward expansion constantly.According to statistics, 80-130 satellite launched in the whole world every year on average, yet there be 2-3 satellite to fail correctly to enter the orbit, and in the satellite of correctly entering the orbit, there be 5-10 again in beginning of lifetime (entering the orbit back preceding 30 days) i.e. inefficacy, wherein, satellite that mechanical breakdown causes lost efficacy accounted for sizable ratio (Tafazoli M.A study of on-orbit spacecraft failures[J] .ActaAstronautica.2009,64:195-205).No. two satellites of prosperous promise of typical example such as China, after launching on October 29th, 2006, though the successful orbit determination of satellite, and subsystems such as the observing and controlling of satellite, rail control all are in good state, yet, because the solar array secondary launches and antenna launches to fail to finish, satellite can't operate as normal, though the scientific research personnel adopts several different methods to succour, still fail to recover its function, these expensive 2,000,000,000 RMB, projected life be 15 years satellite with a useless star.Entirely ineffective or because the satellite abandoned of task termination, resting on space will become space trash, not only take valuable track resources for those, safety that also may other spacecraft of crisis.In order to retrieve a loss as far as possible or to purify orbital environment, it is means that various countries are being studied with the robot for space, with satellite maintenance, life prolong and space trash remove be purpose at rail service technology (Cui Naigang, Wang Ping, Guo Jifeng, Deng. space-orbit service technology development Overview [J]. aerospace journal .2007,28 (4): 33-39.).

At the spacecraft and the space junk of rail service, three characteristics are arranged for great majority: 1) do not have installation to be used for grasping mechanism (handle) and cooperation marker that is used for subsidiary and the characteristic block etc. that mechanical arm is caught; 2) moving state the unknown of target luck may be for rolling under three-axis stabilization, spinning stability or even the runaway condition etc.; 3) there is not direct information interchange between target star and the tracking star.This type of target is called noncooperative target.For robot for space is used for better in the rail service, must solve the autonomous classification of noncooperative target and the key issue of arresting.Simultaneously, in space antagonism, for monitor, destroy, " prisoners of war " enemy satellites, also need to solve the autonomous classification of noncooperative target and catch problem.The identification of noncooperative target is a global problem with catching, and has caused domestic and international researcher's attention.Document (Zhang Shijie, Cao Xibin, Fujian is old. and the monocular vision of relative pose is determined algorithm [J] between non-cooperation spacecraft. the journal .2006 of Institutes Of Technology Of Nanjing, 30 (5): 564-568) proposed a kind of pose measuring method that does not adopt under the cooperation cursor situation, but that hypothesis is identified the shape and the physical dimension of target is known.European Space Agency has designed geostationary orbit restorer ROGER (Robotic Geostationary Orbit Restorer), use rope system to fly net or fly pawl, discarded satellite on the track is arrested (D.A.Smith, C.Martin, M.Kassebom, H.Petersen, A.Shaw, B.Skidmore, D.Smith, H.Stokes, A.Willig. " A mission to preserve the geostationary region ", Advances in Space Research 34 (2004) 1214-1218), wherein the measurement of target has been used and comprised laser ranging, means such as active vision.Document (Thienel J K, Vaneepoel J M, Sanner R M.Accurate state estimation and tracking of a non-cooperative target vehi-cle[C] .AIAA Guidance, Navigation, andControl Confer-ence, Keystone, CO, United States, AIAA 2006-6802 2006:5511-5522.) at the service of Hubble, has proposed the attitude that a kind of non-linear method is used to estimate spacecraft, the line trace of going forward side by side control, but adopted more priori.Document (Inaba N, Oda M, Asano M.Rescuing a stranded satellite in space-experimental robotic captureof non-cooperative satellites[J] .Transactions of the Japan Society for Aero-nautical and SpaceSciences.2006,48 (162): 213-220.) proposed a kind of method that noncooperative target is discerned and caught at rail, but the profile of hypothetical target, size and quality are known.DEOS project (Klaus Landzettel, Alin Albu-that present DLR is carrying out

Bernhard Brunner, et al.ROKVISS:Verification of Advanced Light Weight Robotic Joints and Tele-Presence Concepts for FutureSpace Missions, ICRA2008, Pasadena, California, USA), adopted complicated noncooperative target recognizer, but owing to be subjected to the restriction of satellite-borne processor computing capability, this calculation can't independently be finished on star, but will pass to ground under the image of gathering, under terrestrial operation person's participating in directly, carried out the image processing and calculated object pose by ground installation, measurement result is uploaded on the star again, controller control tracking star is followed the tracks of target on the star, near the operation such as grade.The influence that this method is subjected to propagation delay time, transmission reliability is bigger, and whole system was with instability when time delay was big; And owing to need participating in directly of operating personnel, its " independently " property is not high.Therefore, proposing a kind of non-cooperative target of space robot autonomous classification and catching method, is very necessary and urgent.

Summary of the invention

The objective of the invention is to overcome the deficiencies in the prior art, a kind of non-cooperative target of space robot autonomous classification and catching method are provided, key step comprises based on the image of stereoscopic vision to be handled and the noncooperative target relative position measurement, the autonomous path planning of robot for space target acquistion, and robot for space system coordination control, this method does not require installs luminous sign device or the corner reflector that is used for subsidiary on the target star, also need not to know the physical dimension of target, but directly with the parts of spacecraft self as identifying object, these parts can be the windsurfing supports, antenna holder, apogee engine also can be a satellite-rocket docking ring etc.

Non-cooperative target of space robot autonomous classification of the present invention and catching method comprise the steps:

(1) handles and the noncooperative target relative position measurement based on the image of stereoscopic vision, according to the left and right camera image that collects synchronously, handle in real time, comprise that smothing filtering, edge detect, straight line extracts, target signature is discerned, three-dimensional coupling and 3D reconstruct.Based on the result of reconstruct, calculate the noncooperative target astrology at last for robot for space pedestal and terminal relative position

And attitude

(2) the autonomous path planning of robot for space target acquistion is according to relative pose measurement result, i.e. relative position

And attitude Movement locus---the joint angle Θ in each joint of real-time planning space robot _dAnd angular speed

So that mechanical arm is terminal near also finally catching target;

(3) robot for space system coordination control is coordinated the control in each joint of space manipulator and attitude, the track control of pedestal to carry out, and the position Θ of expectation is followed the tracks of in each joint of control mechanical arm _dAnd speed

Realize the Optimal Control performance of whole system outward.

Described image based on stereoscopic vision is handled and the noncooperative target relative position measurement, may further comprise the steps:

(1) image filtering: respectively left and right sides camera image is carried out filtering,, obtain level and smooth left and right sides camera image to eliminate noise jamming;

(2) edge detects: two width of cloth images are carried out the edge respectively detect, obtain edge features information;

(3) straight line extracts: the image that carries out after the edge detects is carried out the straight line extraction, obtain comprising each straight line of triangular supports in each interior bar straight line information;

(4) identification of noncooperative target satellite A-frame: all the straight line information after extracting, discern six straight lines, and utilize six straight lines of the triangular supports correspondence that identifies, calculate vertex of a triangle corresponding to triangle windsurfing support;

(5) three-dimensional coupling, 3D reconstruct and pose measurement: respectively in left and right magazine 2D information, carry out the 3D reconstruct of each characteristic point according to each summit of the triangular supports that identifies, obtain the 3D coordinate of each summit of triangular supports in world coordinate system; According to the result of 3D reconstruct, further make up and arrest the object coordinate system, and calculate its position and attitude with respect to world coordinate system.

The autonomous path planning of described robot for space target acquistion may further comprise the steps:

(1) the pose deviation is calculated, and the target that stereoscopic vision is measured is with respect to the pose measurement data of pedestal, and converting into target is with respect to the pose of arm end, and judges relative pose deviation e _pAnd e _oWhether less than preset threshold ε _pAnd ε _o, if less than, then closed paw, catch target; Otherwise, other step below carrying out;

(2) prediction of target travel, according to the relative pose deviation, the motion state of real-time estimating target---next position, attitude and linear velocity, angular speed constantly;

(3) the terminal movement velocity planning of robot for space, according to the target state of being predicted, the movement velocity of planning mechanical arm end is to catch target with shortest path;

(4) robot for space is kept away unusual path planning, according to the terminal movement velocity of planning, adopts the processing method of unusual avoidance, the movement locus of each joint angle of planning mechanical arm.

Described robot for space system coordination control makes " mechanical arm controller " and " pedestal controller " collaborative work, realizes the optimization of whole system control performance.The motion of mechanical arm control system control mechanical arm, and estimate the reaction of manipulator motion for satellite, pass to the AOCS of rail control system of satellite pedestal with the form of moment or angular momentum, by AOCS this reaction is compensated, make the attitude of satellite keep stable, the attitude measurement element is with attitude state of a control---attitude angle simultaneously, attitude angle speed, flywheel and jet state are passed to mechanical arm control system again, mechanical arm system is judged motion planning according to these states, whether plan again with decision, if attitude angle, angular speed exceeds default scope, the path that planning is described is bad, needs planning again.The control system of pedestal is also carried out feedforward compensation according to the manipulator motion that estimates to the reaction of satellite except carry out conventional FEEDBACK CONTROL according to the attitude measurement state;

Described image filtering adopts median filtering algorithm, described edge to detect and adopts Canny algorithm, described straight line extraction employing Hough mapping algorithm.Medium filtering is for given n numerical value { a ₁, a ₂..., a _n, with they order arrangements by size, when n was odd number, that numerical value that is positioned at the centre position was called the intermediate value of this n numerical value.When n was even number, the mean value that is positioned at two numerical value in centre position was called the intermediate value of this n numerical value, and note is made med{a ₁, d ₂..., a _n, image is through behind the medium filtering, and the output of certain pixel equals the intermediate value of each pixel grey scale in this pixel field; The Canny edge detection algorithm with the Gaussian filter smoothed image, with the finite difference of single order local derviation assign to compute gradient amplitude and direction, to gradient magnitude use non-maximum suppress, with the detection of dual threshold algorithm be connected the edge; The Hough mapping algorithm with the straight line among the image space x-y, is for conversion into a bit of parameter space with the duality of point-line;

The identification of described noncooperative target satellite A-frame comprises initial acquisition and two key steps of real-time tracking, and initial acquisition is by being provided with two initial reference point p to the left and right camera image that passes to ground under the telemetering channel _R1, p _R2, spaceborne program can be finished the automatic identification to the windsurfing support; After initial acquisition was set up, spaceborne program changed tracing mode immediately over to, later on can be automatically according to image, and the zone at real-time tracking windsurfing support place.Reference point p _R1For in the triangle more arbitrarily, p _R2For triangle outer and be positioned at that solar battery array lists more arbitrarily, p _R1, p _R2Directly in the remote measurement image, choose by mouse, do not have strict especially requirement;

Described image is handled and the noncooperative target relative position measurement, the image coordinate that it is characterized in that the triangular supports summit that described three-dimensional coupling, 3D reconstruct and the utilization of pose measurement step identify respectively from left and right camera image, adopt least square method that three-dimensional projection model is calculated, obtain the 3D coordinate of triangular supports summit in world coordinate system, and, calculate position, the attitude of target-based coordinate system with respect to world coordinate system according to this coordinate establishing target coordinate system of 3;

The autonomous path planning of described robot for space target acquistion, it is characterized in that the prediction steps of described target travel at first sets up the state equation of the motion of target, adopt next relative position, attitude and relative linear velocity, the angular speed constantly of expansion Kalman wave filter target of prediction satellite then;

The autonomous path planning of described robot for space target acquistion is characterized in that the movement velocity of the terminal movement velocity planning step of described robot for space according to following formula planning mechanical arm end:

$\left[\begin{array}{c}{}^{E}v_{\mathrm{ed}}\\ {}^E\mathrm{\ω}_{\mathrm{ed}}\end{array}\right]=K\left[\begin{array}{c}{}^{E}r_{\mathrm{eh}}\\ \mathrm{\ΔO}\end{array}\right]+\left[\begin{array}{c}{}^{E}v_h\\ {}^E\mathrm{\ω}_h\end{array}\right]---\left(1\right)$

In the following formula, ^Ev _Ed, ^Eω _EdBe respectively the terminal linear velocity and the angular speed of planning; K is a gain matrix, is diagonal matrix, is used to limit the terminal maximal rate of planning; ^Er _Eh, ^Ev _h, ^Eω _hBe respectively next relative position, linear velocity and the angular speed constantly of target of prediction, the expression in the terminal coordinate system of mechanical arm, Δ O is an attitude error, is calculated as follows:

$\mathrm{\ΔO}=\frac{1}{2}(n_e\×n_h+o_e\×o_h+a_e\×a_h)=\frac{1}{2}\left[\begin{array}{c}{}^{E}A_H\left(\mathrm{2,3}\right)-{}^{E}A_H\left(\mathrm{3,2}\right)\\ -{}^{E}A_H\left(\mathrm{1,3}\right)+{}^{E}A_H\left(\mathrm{3,1}\right)\\ {}^{E}A_H\left(\mathrm{1,2}\right)-{}^{E}A_H\left(\mathrm{2,1}\right)\end{array}\right]---\left(2\right)$

Wherein, [n _e, o _e, a _e] and [n _h, o _h, a _h] be respectively the posture changing matrix of terminal coordinate system of mechanical arm and target-based coordinate system, and ^EA _HBe the handle measured by the trick camera attitude matrix with respect to terminal coordinate system.

The autonomous path planning of described robot for space target acquistion is characterized in that described robot for space keeps away unusual path planning the unusual avoidance of the dynamics of robot for space is converted into the unusual avoidance of real-time kinematics, promptly

$\left[\begin{array}{c}{}^{E}v_{\mathrm{ed}}\\ {}^E\mathrm{\ω}_{\mathrm{ed}}\end{array}\right]-\left({}^E\hat{J}_b\right)\left({}^E\mathrm{\ω}_0\right)=\left({}^E\hat{J}_m\right)\·{\stackrel{\·}{\mathrm{\Θ}}}_d---\left(3\right)$

The following formula left side is the movement velocity (expression in base coordinate system) of mechanical arm end with respect to pedestal, thereby can be written as:

$\left[\begin{array}{c}{}^E{v}_e^0\\ {}^E{\mathrm{\ω}}_e^0\end{array}\right]={}^E\hat{J}_m{\stackrel{\·}{\mathrm{\Θ}}}_d---\left(4\right)$

^Ev _e ⁰, ^Eω _e ⁰Represent of linear velocity, angular speed the expression in base coordinate system of mechanical arm end respectively with respect to pedestal;

Be the common Jacobian matrix of mechanical arm, thus, find the solution the joint angle speed of expectation according to (4)

The present invention compared with prior art has following advantage: noncooperative target autonomous classification and the pose measurement that adopt (1), do not require luminous sign device or the corner reflector that is used for subsidiary is installed on the target star, also need not to know the physical dimension of target, the parts of Direct Recognition spacecraft self, and calculate relative pose; (2) the autonomous paths planning method of robot for space target acquistion of Cai Yonging, can real-time estimate target motion state, independently avoid interference and the autonomous path of adjusting planning that kinematics and dynamics are unusual, the real-time estimate manipulator motion produces pedestal; (3) control method for coordinating of Cai Yonging had both been considered the control performance of whole robot for space system, had adapted to the control ability of current satellite-borne processor again.

Description of drawings

Fig. 1 is that the typical space robot is in the rail service procedure;

Fig. 2 is that robot for space system and noncooperative target are caught schematic diagram at rail;

Fig. 3 is non-cooperative target of space robot autonomous classification and catching method flow chart;

Fig. 4 is based on the image of stereoscopic vision and handles and noncooperative target relative position measurement algorithm flow;

Fig. 5 is the virtual left camera image (512 * 512 black and white) that simulation generates;

Fig. 6 is the virtual right camera image (512 * 512 black and white) that simulation generates;

Fig. 7 is the windsurfing support of the noncooperative target that identified;

Fig. 8 is the definition of intersection reference frame and capture point coordinate system;

Fig. 9 is the autonomous path planning algorithm flow process that robot for space is caught noncooperative target;

Figure 10 is the control method for coordinating structure that robot for space is caught noncooperative target.

The specific embodiment

One, the main flow process of non-cooperative target of space robot autonomous classification and catching method

The typical space robot promptly utilizes the robot for space system in the rail service idea as shown in Figure 1, to fault satellites follow the tracks of, approaching, catch, dock, operation such as maintenance.Be broadly divided into following several stages in the rail service procedure:

(a) remote follow the tracks of near (＞300m): the position of robot for space system about from self orbit maneuver to distance objective satellite 300m;

(b) middle distance is followed the tracks of approaching: the robot for space system is from the motor-driven position to about distance objective satellite 15m of 300m;

(c) closely intersection and stop: the robot for space system is motor-driven position to about 1m from the position about 15m, and stops relatively, makes target satellite be in the working space of mechanical arm;

(d) arrest and dock at rail: utilize space manipulator capture target satellite, and itself and robot for space pedestal are docking together;

(e) keep in repair or carry out the operation that leaves the right or normal track at rail, be about to target and bring earth atmosphere or grave track into: utilize mechanical arm that target satellite is carried out perhaps target being brought into earth atmosphere or grave track in the rail maintenance.

Suppose to treat that the service object is the fault satellites that a windsurfing does not launch, install in advance on it and be used for the marker of vision measurement and the handle that is used to catch, be noncooperative target.Designed robot for space is made up of a flight pedestal and manipulator in the rail service system, as shown in Figure 2.Wherein, flight has been installed the target measurement system on the pedestal---pedestal stereoscopic vision, docking mechanism, rail control system etc., space manipulator by the 6DOF mechanical arm, arrest paw and terminal stereoscopic vision is formed.For the implementation space robot system to fault satellites follow the tracks of, approaching, arrest, operation such as maintenance, adopt as shown in Figure 3 noncooperative target autonomous classification and catching method, key step comprises:

(1) handles and the noncooperative target relative position measurement based on the image of stereoscopic vision, according to the left and right camera image that collects synchronously, handle in real time, comprise that smothing filtering, edge detect, straight line extracts, target signature is discerned, three-dimensional coupling and 3D reconstruct.Based on the result of reconstruct, calculate the noncooperative target astrology at last for robot for space pedestal and terminal relative position

And attitude

(2) the autonomous path planning of robot for space target acquistion, according to the relative pose measurement result, and relative position And attitude

Movement locus---the joint angle Θ in each joint of real-time planning space robot _dAnd angular speed So that mechanical arm is terminal near also finally catching target;

(3) robot for space system coordination control is coordinated the control in each joint of space manipulator and attitude, the track control of pedestal to carry out, and the position Θ of expectation is followed the tracks of in each joint of control mechanical arm _dAnd speed

Realize the Optimal Control performance of whole system outward.

Two, handle and the noncooperative target relative position measurement based on the image of stereoscopic vision

Image processing based on stereoscopic vision is handled the left and right camera image that collects synchronously with noncooperative target relative position measurement algorithm, and calculates position and the attitude of target with respect to pedestal, and flow process as shown in Figure 4.With as shown in Figure 5 and Figure 6 left and right camera analog image is example, and key step is as follows:

(1) image filtering: respectively left and right sides camera image is carried out filtering,, obtain level and smooth left and right sides camera image to eliminate noise jamming

The actual image that obtains generally all contains noise, and the present invention adopts median filtering algorithm to carry out image smoothing.Medium filtering is a kind of Nonlinear Processing method that suppresses noise.For given n numerical value { a ₁, a ₂..., a _n, with they order arrangements by size.When n was odd number, that numerical value that is positioned at the centre position was called the intermediate value of this n numerical value.When n was even number, the mean value that is positioned at two numerical value in centre position was called the intermediate value of this n numerical value, and note is made med{a ₁, a ₂..., a _n.Image is through behind the medium filtering, and the output of certain pixel equals the intermediate value of each pixel grey scale in this pixel field.The method computing of medium filtering is simple, is easy to realize, and can protects the border preferably.

(2) edge detects: two width of cloth images are carried out the edge respectively detect, obtain edge features information

Edge (edge) is meant the most significant part of image local Strength Changes.The edge mainly was present between target and target, target and background, zone and regional (comprising different color), was that image is cut apart, the important foundation of graphical analyses such as textural characteristics and shape facility.This paper adopts the Canny algorithm that image is carried out rim detection.The step of Canny edge detection algorithm is as follows: (a) use the Gaussian filter smoothed image; (b) with the finite difference of single order local derviation the assign to amplitude and the direction of compute gradient; (c) gradient magnitude being used non-maximum suppresses; (d) usefulness dual threshold algorithm detects and is connected the edge.

(3) straight line extracts: the image that carries out after the edge detects is carried out the straight line extraction, obtain comprising each straight line of triangular supports in each interior bar straight line information

After utilizing the Canny algorithm to detect the image edge, adopt the Hough conversion to extract straight line again.The basic thought of hough transform is to utilize a little-duality of line (point-straight line or point-curve), i.e. straight line among the image space x-y, through become after the Hough conversion parameter space a bit.

The step of Hough mapping algorithm is as follows: (a) quantization parameter space suitably; (b) each unit of supposition parameter space all is an accumulator, is accumulator initialization zero; (c) to the every bit of image space, on the accumulator of the parametric equation correspondence that it satisfied, add 1; (d) parameter of the corresponding model of the maximum of accumulator array.

(4) identification of A-frame: all the straight line information after extracting, identification is corresponding to six straight lines of triangle windsurfing support

After adopting the Hough conversion, can extract many straight lines, comprise profile, solar cell array and the solar array support of target satellite.Because the windsurfing support has obvious characteristics, so be identifying object with the windsurfing support.Comforming and identify six straight lines that belong to the windsurfing support in the multi straight, is the key that this paper studies.Owing to, do not have that other any prioris---comprise the gradient on the leg-of-mutton length of side, interior angle, each limit etc., therefore, the difficulty of identification is well imagined except known windsurfing support is the triangle.Can pass to ground by under the telemetering channel at this supposition image, terrestrial operation person is according to two initial reference point p of this image setting _R1, p _R2, spaceborne program can be finished the automatic identification to the windsurfing support, changes tracing mode immediately over to, later on can be automatically according to image, and the zone at real-time tracking windsurfing support place.Reference point p _R1For in the triangle more arbitrarily, p _R2For triangle outer and be positioned at that solar battery array lists more arbitrarily, p _R1, p _R2Directly in the remote measurement image, choose by mouse, do not have strict especially requirement.The principle of identification and result are as shown in Figure 7.

(5) intersection point Feature Extraction: utilize six straight lines of the triangular supports correspondence that identifies, calculate vertex of a triangle

(6) three-dimensional coupling, 3D reconstruct: respectively in left and right magazine 2D information, carry out the 3D reconstruct of each characteristic point according to each summit of the triangular supports that identifies, obtain the 3D coordinate of each summit of triangular supports in world coordinate system, promptly ^WP ₁～ ^WP ₆:

At first, camera imaging can be expressed as with pin-hole model

$\mathrm{\λ}\left[\begin{array}{c}u\\ v\\ 1\end{array}\right]=C\left[\begin{array}{c}{X}_W\\ Y_W\\ Z_W\\ 1\end{array}\right]---\left(5\right)$

Wherein, ^WP=[X _W, Y _W, Z _W] ^TBe the coordinate of some P in space in world coordinate system, (u v) is to be the image coordinate of unit with the pixel, and C is the projective transformation matrix of camera, is determined by its inside and outside parameter.3D based on stereoscopic vision rebuilds principle as shown in Figure 7, supposes i characteristic point P _iPosition in world coordinate system is ^WP _i=[X _WiY _WiZ _Wi] ^T, be projected in left and right sides image of camera coordinate and be respectively p _Li=[u _Liv _Li] ^TAnd p _Ri=[u _Riv _Ri] ^TAccording to (5),, following relation is arranged for left and right camera:

$\left[\begin{array}{c}{\mathrm{\λ}}_L{u}_{L_i}\\ {\mathrm{\λ}}_L{v}_{\mathrm{Li}}\\ {\mathrm{\λ}}_L\end{array}\right]=C_L\left[\begin{array}{c}{X}_{\mathrm{wi}}\\ Y_{\mathrm{wi}}\\ Z_{\mathrm{wi}}\\ 1\end{array}\right]=\left[\begin{array}{cccc}1& {\mathrm{\α}}_1& {\mathrm{\α}}_2& {\mathrm{\α}}_3\\ {\mathrm{\α}}_4& {\mathrm{\α}}_5& {\mathrm{\α}}_6& {\mathrm{\α}}_7\\ {\mathrm{\α}}_8& {\mathrm{\α}}_9& {\mathrm{\α}}_{10}& {\mathrm{\α}}_{11}\end{array}\right]\left[\begin{array}{c}{X}_{\mathrm{wi}}\\ Y_{\mathrm{wi}}\\ Z_{\mathrm{wi}}\\ 1\end{array}\right]---\left(6\right)$

$\left[\begin{array}{c}{\mathrm{\λ}}_R{u}_{\mathrm{Ri}}\\ {\mathrm{\λ}}_R{v}_{\mathrm{Ri}}\\ {\mathrm{\λ}}_R\end{array}\right]=C_R\left[\begin{array}{c}{X}_{\mathrm{wi}}\\ Y_{\mathrm{wi}}\\ Z_{\mathrm{wi}}\\ 1\end{array}\right]=\left[\begin{array}{cccc}1& {\mathrm{\β}}_1& {\mathrm{\β}}_2& {\mathrm{\β}}_3\\ {\mathrm{\β}}_4& {\mathrm{\β}}_5& {\mathrm{\β}}_6& {\mathrm{\β}}_7\\ {\mathrm{\β}}_8& {\mathrm{\β}}_9& {\mathrm{\β}}_{10}& {\mathrm{\β}}_{11}\end{array}\right]\left[\begin{array}{c}{X}_{\mathrm{wi}}\\ Y_{\mathrm{wi}}\\ Z_{\mathrm{wi}}\\ 1\end{array}\right]---\left(7\right)$

Matrix C _L, C _RBe respectively the transformation matrix of left and right sides camera, α _i, β _iElement for correspondence.Formula (6) and (7) are carried out abbreviation, have:

$\left[\begin{array}{ccc}1-{\mathrm{\α}}_8{u}_{\mathrm{Li}}& {\mathrm{\α}}_1-{\mathrm{\α}}_9{u}_{\mathrm{Li}}& {\mathrm{\α}}_2-{\mathrm{\α}}_{10}{u}_{\mathrm{Li}}\\ {\mathrm{\α}}_4-{\mathrm{\α}}_8{v}_{\mathrm{Li}}& {\mathrm{\α}}_5-{\mathrm{\α}}_9{v}_{\mathrm{Li}}& {\mathrm{\α}}_6-{\mathrm{\α}}_{10}{v}_{\mathrm{Li}}\\ 1-{\mathrm{\β}}_8{u}_{\mathrm{Ri}}& {\mathrm{\β}}_1-{\mathrm{\β}}_9{u}_{\mathrm{Ri}}& {\mathrm{\β}}_2-{\mathrm{\β}}_{10}{u}_{\mathrm{Ri}}\\ {\mathrm{\β}}_4-{\mathrm{\β}}_8{v}_{\mathrm{Ri}}& {\mathrm{\β}}_5-{\mathrm{\β}}_9{v}_{\mathrm{Ri}}& {\mathrm{\β}}_6-{\mathrm{\β}}_{10}{v}_{\mathrm{Ri}}\end{array}\right]\left[\begin{array}{c}{X}_{\mathrm{wi}}\\ Y_{\mathrm{wi}}\\ Z_{\mathrm{wi}}\end{array}\right]=\left[\begin{array}{c}{\mathrm{\α}}_{11}{u}_{\mathrm{Li}}-{\mathrm{\α}}_3\\ {\mathrm{\α}}_{11}{v}_{\mathrm{Li}}-{\mathrm{\α}}_7\\ {\mathrm{\β}}_{11}{u}_{\mathrm{Ri}}-{\mathrm{\β}}_3\\ {\mathrm{\β}}_{11}{v}_{\mathrm{Ri}}-{\mathrm{\β}}_7\end{array}\right]---\left(8\right)$

Write as the form of matrix:

K· ^WP _i＝U (9)

Formula (8) has four equations, three unknown numbers, is transcendental equation, and available least square method is found the solution:

^WP _i＝(K ^TK) ^-1K ^TU (10)

Said process is the 3D restructuring procedure based on stereoscopic vision.

(7) measurement of target location, attitude:, further make up and arrest the object coordinate system, and calculate it with respect to world coordinate system according to the result of 3D reconstruct---follow the trail of position and attitude that star is measured coordinate system.

According to six characteristic points that identified, can set up " intersection reference frame " ∑ of target _R2" capture point coordinate system " ∑ _Cap, as shown in Figure 8.At first, three of carriage center point (Q ₁～Q ₃) three-dimensional coordinate determine by following several formulas:

${}^{W}Q_1=\frac{{}^{W}P_1+{}^{W}P_2}{2}---\left(11\right)$

${}^{W}Q_2=\frac{{}^{W}P_3+{}^{W}P_4}{2}---\left(12\right)$

${}^{W}Q_3=\frac{{}^{W}P_5+{}^{W}P_6}{2}---\left(13\right)$

Mid point then

${}^{W}M=\frac{{}^{W}Q_1+{}^{W}Q_2}{2}---\left(14\right)$

The coordinate system ∑ _R2Initial point be

$r_{R2}={}^{W}O_t=\frac{{}^{W}Q_3+{}^{W}M}{2}---\left(15\right)$

The coordinate system ∑ _R2X-axis be unit vector:

$n_{R2}=\frac{Q_3{Q}_1\×Q_3{Q}_2}{\left|\right|Q_3{Q}_1\×Q_3{Q}_2\left|\right|}---\left(16\right)$

The coordinate system ∑ _R2Y-axis be unit vector:

$o_{R2}=\frac{Q_{3}M}{\left|\right|Q_{3}M\left|\right|}---\left(17\right)$

The coordinate system ∑ _R2The Z axle determine by the right-hand rule:

a _R2＝n _R2×o _R2 (18)

Coordinate system ∑ then _R2With respect to intersection referential ∑ _R1The attitude spin matrix be

^R1A _R2＝[n _R2?o _R2?a _R2] (19)

The attitude Eulerian angles can be passed through the attitude spin matrix ^R1A _R2Obtain, the relative position that convolution (15) is obtained, so far, the relative position and the attitude of target are all measured.Similarly, " capture point coordinate system " ∑ _CapWith respect to " intersection referential " ∑ _R1Position, attitude also can calculate.

Three, the autonomous path planning of robot for space target acquistion

Above-mentioned image is handled with the pose measurement algorithm and has been provided " capture point coordinate system " ∑ _CapWith respect to " intersection referential " ∑ _R1Position, attitude, autonomous planning algorithm at first is converted into this measurement result earlier " capture point coordinate system " ∑ _CapWith respect to position, the attitude of " the terminal coordinate system of mechanical arm ", then according to this result, the motion of planning space robot in real time is finally to catch target.Mainly comprise as shown in Figure 9, comprising: the measurement of trick camera, the calculating of pose deviation, the prediction of target travel, the terminal movement velocity planning of robot for space, robot for space are kept away unusual path planning etc.At first, judge relative pose deviation e according to the trick measurement data _pAnd e _oWhether less than preset threshold ε _pAnd ε _o(be capture region, in 10mm, the three-axis attitude deviation is about 1 ° as three relative positions) is if less than, then closed paw, catch target; Otherwise, then according to the relative pose deviation, the motion state of real-time estimating target, and results estimated is reacted in the planning of end of arm speed, terminal to guarantee mechanical arm constantly towards nearest direction (straight line) convergence target, target is to the last caught in the terminal independently motion of tracking target of mechanical arm.After cooking up terminal movement velocity, promptly call autonomous unusual backoff algorithm, to resolve the expectation angular speed in joint.

Adopt next relative position, attitude and relative linear velocity, the angular speed constantly of Kalman wave filter target of prediction, based on prediction result, the terminal movement velocity of expectation is planned by following formula:

$\left[\begin{array}{c}{}^{E}v_{\mathrm{ed}}\\ {}^E\mathrm{\ω}_{\mathrm{ed}}\end{array}\right]=K\left[\begin{array}{c}{}^{E}r_{\mathrm{eh}}\\ \mathrm{\ΔO}\end{array}\right]+\left[\begin{array}{c}{}^{E}v_h\\ {}^E\mathrm{\ω}_h\end{array}\right]---\left(20\right)$

In the following formula, ^Ev _h, ^Eω _hFor the absolute movement speed of target at ∑ _EIn expression, Δ O is attitude error (being the attitude of handle coordinate system with respect to the terminal coordinate system of mechanical arm), is calculated as follows:

$\mathrm{\ΔO}=\frac{1}{2}(n_e\×n_h+o_e\×o_h+a_e\×a_h)=\frac{1}{2}\left[\begin{array}{c}{}^{E}A_H\left(\mathrm{2,3}\right)-{}^{E}A_H\left(\mathrm{3,2}\right)\\ -{}^{E}A_H\left(\mathrm{1,3}\right)+{}^{E}A_H\left(\mathrm{3,1}\right)\\ {}^{E}A_H\left(\mathrm{1,2}\right)-{}^{E}A_H\left(\mathrm{2,1}\right)\end{array}\right]---\left(21\right)$

Wherein, [n _e, o _e, a _e] and [n _h, o _h, a _h] be respectively ∑ _EAnd ∑ _HPosture changing matrix (with respect to inertial system), and ^EA _HBe the handle measured by the trick camera attitude matrix with respect to terminal coordinate system.According to relative motion principle, the absolute movement speed of target is obtained by following formula:

${}^{E}v_h={}^E{v}_h^e+{}^{E}v_e+{}^E\mathrm{\ω}_e\×{}^{E}r_{\mathrm{eh}}---\left(22\right)$

${}^E\mathrm{\ω}_h={}^E{\mathrm{\ω}}_h^e+{}^E\mathrm{\ω}_e---\left(23\right)$

Wherein, ^Ev _h ^eRepresent the linear velocity of the handle of target star with respect to the mechanical arm end, the expression in the coordinate system endways, that is:

${}^E{v}_h^e={[\stackrel{\·}{X},\stackrel{\·}{Y},\stackrel{\·}{Z}]}^T---\left(24\right)$

${}^E{\mathrm{\ω}}_h^e=\left[\begin{array}{ccc}0& -s_{\mathrm{\α}}& c_{\mathrm{\α}}{c}_{\mathrm{\β}}\\ 0& c_{\mathrm{\α}}& s_{\mathrm{\α}}{c}_{\mathrm{\β}}\\ 1& 0& -s_{\mathrm{\β}}\end{array}\right]\left[\begin{array}{c}\stackrel{\·}{\mathrm{\α}}\\ \stackrel{\·}{\mathrm{\β}}\\ \stackrel{\·}{\mathrm{\γ}}\end{array}\right]---\left(25\right)$

According to formula (22) and (23), the movement velocity of target

$\left[\begin{array}{c}{}^{E}v_h\\ {}^E\mathrm{\ω}_h\end{array}\right]=\left[\begin{array}{c}{}^E{v}_h^e\\ {}^E{\mathrm{\ω}}_h^e\end{array}\right]+\left[\begin{array}{cc}{I}_{3\×3}& -{}^{E}r_{\mathrm{eh}}\\ O_{3\×3}& I_{3\×3}\end{array}\right]\left[\begin{array}{c}{}^{E}v_e\\ {}^E\mathrm{\ω}_e\end{array}\right]---\left(26\right)$

Differential kinematics equation according to the free-floating robot for space has

$\left[\begin{array}{c}{}^{E}v_e\\ {}^E\mathrm{\ω}_e\end{array}\right]=\left({}^{E}J_g\right)\stackrel{\·}{\mathrm{\Θ}}---\left(27\right)$

Wherein, ^EJ _gBe the terminal generalized Jacobian of mechanical arm,

Be the current angular speed of mechanical arm,, have formula (27) substitution (26)

$\left[\begin{array}{c}{}^{E}v_h\\ {}^E\mathrm{\ω}_h\end{array}\right]=\left[\begin{array}{c}{}^E{v}_h^e\\ {}^E{\mathrm{\ω}}_h^e\end{array}\right]+\left[\begin{array}{cc}{I}_{3\×3}& -{}^{E}r_{\mathrm{eh}}\\ O_{3\×3}& I_{3\×3}\end{array}\right]\left({}^{E}J_g\right)\stackrel{\·}{\mathrm{\Θ}}---\left(28\right)$

At last, according to formula (28) and (1), can plan the terminal movement velocity of mechanical arm by following formula:

$\left[\begin{array}{c}{}^{E}v_{\mathrm{ed}}\\ {}^E\mathrm{\ω}_{\mathrm{ed}}\end{array}\right]=K\left[\begin{array}{c}{}^{E}r_{\mathrm{eh}}\\ \mathrm{\ΔO}\end{array}\right]+\left[\begin{array}{c}{}^E{v}_h^e\\ {}^E{\mathrm{\ω}}_h^e\end{array}\right]+\left[\begin{array}{cc}{I}_{3\×3}& -{}^{E}r_{\mathrm{eh}}\\ O_{3\×3}& I_{3\×3}\end{array}\right]\left({}^{E}J_g\right)\stackrel{\·}{\mathrm{\Θ}}---\left(29\right)$

On the other hand, the differential kinematics equation of robot for space can be expressed as

$\left[\begin{array}{c}{}^{E}v_{\mathrm{ed}}\\ {}^E\mathrm{\ω}_{\mathrm{ed}}\end{array}\right]=\left({}^E\hat{J}_b\right)\left({}^E\mathrm{\ω}_0\right)+\left({}^E\hat{J}_m\right)\·{\stackrel{\·}{\mathrm{\Θ}}}_d---\left(30\right)$

Through after the suitable modification, document (Xu Wenfu, refined, Liu Yu, Lee becomes, strong context. a kind of new unusual backoff algorithm of PUMA type machine people [J]. automation journal .2008,34 (6): the unusual backoff algorithm of introducing 670-675) of free-floating robot for space can be used for the unusual avoidance of autonomous path planning.Wrist motion speed of representing in the coordinate system at first, endways and the pass between the terminal movement velocity are:

${}^{E}v_w={}^{E}v_e-{}^E\mathrm{\ω}_w\×\left[\begin{array}{c}0\\ 0\\ d_6\end{array}\right]={}^{E}v_e-\left[\begin{array}{c}{d}_6\left({}^E\mathrm{\ω}_{\mathrm{wy}}\right)\\ -d_6\left({}^E\mathrm{\ω}_{\mathrm{wx}}\right)\\ 0\end{array}\right]={}^{E}v_e-\left[\begin{array}{c}{d}_6\left({}^E\mathrm{\ω}_{\mathrm{ey}}\right)\\ -d_6\left({}^E\mathrm{\ω}_{\mathrm{ex}}\right)\\ 0\end{array}\right]---\left(31\right)$

^Eω _w＝ ^Eω _e (32)

Secondly, mechanical arm wrist Jacobian matrix:

${}^E\hat{J}_W=\left[\begin{array}{cc}{}^E\hat{J}_{11}& O\\ {}^E\hat{J}_{21}& {}^E\hat{J}_{22}\end{array}\right]---\left(33\right)$

Wherein

${}^E\hat{J}_{11}=\left[\begin{array}{ccc}({\hat{d}}_4{s}_{23}+{\hat{a}}_2{c}_2)(s_4{c}_5{c}_6+c_4{s}_6)& (c_4{c}_5{c}_6-s_4{s}_6)({\hat{d}}_4+{\hat{a}}_2{s}_3)+{\hat{a}}_2{c}_3{s}_5{c}_6& {\hat{d}}_4(c_4{c}_5{c}_6-s_4{s}_6)\\ ({\hat{d}}_4{s}_{23}+{\hat{a}}_2{c}_2)(-s_4{c}_5{s}_6+c_4{c}_6)& -(c_4{c}_5{s}_6+s_4{c}_6)({\hat{d}}_4+{\hat{a}}_2{s}_3)-{\hat{a}}_2{c}_3{s}_5{s}_6& -{\hat{d}}_4(c_4{c}_5{s}_6+s_4{c}_6)\\ ({\hat{d}}_4{s}_{23}+{\hat{a}}_2{c}_2)s_4{s}_5& c_4{s}_5({\hat{d}}_4+{\hat{a}}_2{s}_3)-{\hat{a}}_2{c}_3{c}_5& {\hat{d}}_4{c}_4{s}_5\end{array}\right]---\left(34\right)$

${}^E\hat{J}_{21}=\left[\begin{array}{ccc}-s_{23}(c_4{c}_5{c}_6-s_4{s}_6)-c_{23}{s}_5{c}_6& s_4{c}_5{c}_6+c_4{s}_6& s_4{c}_5{c}_6+c_4{s}_6\\ s_{23}(c_4{c}_5{s}_6+s_4{c}_6)+c_{23}{s}_5{s}_6& -s_4{c}_5{s}_6+c_4{c}_6& -s_4{c}_5{s}_6+c_4{c}_6\\ -s_{23}{c}_4{s}_5+c_{23}{c}_5& s_4{s}_5& s_4{s}_5\end{array}\right]---\left(35\right)$

${}^E\hat{J}_{22}=\left[\begin{array}{ccc}-s_5{c}_6& s_6& 0\\ s_5{s}_6& c_6& 0\\ c_5& 0& 1\end{array}\right]---\left(36\right)$

Can derive:

${}^{6}A_3=\left[\begin{array}{ccc}{c}_4{c}_5{c}_6-s_4{s}_6& s_4{c}_5{c}_6+c_4{s}_6& -s_5{c}_6\\ -(c_4{c}_5{s}_6+s_4{c}_6)& -s_4{c}_5{s}_6+c_4{c}_6& s_5{s}_6\\ c_4{s}_5& s_4{s}_5& c_5\end{array}\right]=\left[\begin{array}{ccc}n& o& a\end{array}\right]---\left(37\right)$

Then

${}^E\hat{J}_{11}=\left({}^{6}A_3\right)\left({}^3\hat{J}_{11}\right)---\left(38\right)$

In like manner

${}^{6}A_5=\left[\begin{array}{ccc}{c}_6& s_6& 0\\ -s_6& c_6& 0\\ 0& 0& 1\end{array}\right]---\left(39\right)$

${}^E\hat{J}_{22}=\left[\begin{array}{ccc}-s_5{c}_6& s_6& 0\\ s_5{s}_6& c_6& 0\\ c_5& 0& 1\end{array}\right]=\left[\begin{array}{ccc}{c}_6& s_6& 0\\ -s_6& c_6& 0\\ 0& 0& 1\end{array}\right]\left[\begin{array}{ccc}-s_5& 0& 0\\ 0& 1& 0\\ c_5& 0& 1\end{array}\right]=\left({}^{6}A_5\right)\left({}^5\hat{J}_{22}\right)---\left(40\right)$

Thus, can with document (Xu Wenfu, refined, Liu Yu, Lee become, strong context. a kind of new unusual backoff algorithm of PUMA type machine people [J]. automation journal .2008,34 (6): unusual backoff algorithm 670-675) is used for the autonomous path planning of target acquistion.

Four, robot for space system coordination control

When the pedestal attitude angle of expectation and after joint of mechanical arm angle trajectory planning comes out, need the design corresponding control strategies, the track that control pedestal attitude and joint of mechanical arm angle tracking are planned.Traditional mode is that pedestal and mechanical arm two parts are independently controlled separately, i.e. independent design controller is separately controlled by following PD control law respectively as pedestal attitude and joint angle:

$\left\{\begin{array}{c}{\mathrm{\τ}}_b=k_{p0}({\mathrm{\θ}}_{0d}-{\mathrm{\θ}}_0)+k_{d0}({\mathrm{\ω}}_{0d}-{\mathrm{\ω}}_0)\\ {\mathrm{\τ}}_m=K_p({\mathrm{\Θ}}_d-\mathrm{\Θ})+K_d({\stackrel{\·}{\mathrm{\Θ}}}_d-\stackrel{\·}{\mathrm{\Θ}})\end{array}\right.---\left(41\right)$

Yet, under this control mode, do things in his own way between " mechanical arm control system " and " pedestal control system ", be difficult to realize the optimization of whole system control performance, motion as the mechanical arm planned may produce the interference that exceeds the pedestal control ability, cause the pedestal attitude controller can't realize that the attitude of expecting changes, and influences the location of mechanical arm end conversely again.Simultaneously, under existence conditions, the controller that design can be controlled pedestal attitude and joint of mechanical arm angle simultaneously is impossible, therefore, this paper considers to adopt the method for coordinating control, though still be divided into " mechanical arm control system " and " pedestal control system " from hardware, their control behavior is " coordinating mutually ".The wherein motion of mechanical arm control system control mechanical arm, and estimate the reaction of manipulator motion for satellite, pass to the AOCS of rail control system of satellite pedestal with the form of moment or angular momentum, by AOCS this reaction is compensated, make the attitude of satellite keep stable, the attitude measurement element is passed to mechanical arm control system again with the attitude state of a control simultaneously, mechanical arm system is judged motion planning according to these states, whether plan again with decision, if attitude angle, angular speed exceed default scope, the path that planning is described is bad, needs planning again.The control system of pedestal is also carried out feedforward compensation according to the manipulator motion that estimates to the reaction of satellite except carry out conventional FEEDBACK CONTROL according to the attitude measurement state.So just realized the coordination control of the attitude of satellite and manipulator motion, this is a kind of control method of distributed and collaboration type.The thought of coordinating control is shown in the empty frame of Figure 10 Smalt.

The content that is not described in detail in the specification of the present invention belongs to this area professional and technical personnel's known prior art.

Claims (10)

1. non-cooperative target of space robot autonomous classification and catching method is characterized in that may further comprise the steps:

(1) handles and noncooperative target relative position measurement 1 based on the image of stereoscopic vision, according to the left and right camera image that collects synchronously, handle in real time, comprise that smothing filtering, edge detect, straight line extracts, target signature is discerned, three-dimensional coupling and 3D reconstruct.Based on the result of reconstruct, calculate the noncooperative target astrology at last for robot for space pedestal and terminal relative position

And attitude

(2) the autonomous path planning 2 of robot for space target acquistion is according to relative pose measurement result, i.e. relative position

And attitude

Movement locus---the joint angle in each joint of real-time planning space robot

And angular speed

So that mechanical arm is terminal near also finally catching target;

(3) robot for space system coordination control 3, and the control in each joint of space manipulator and attitude, the track control of pedestal are coordinated to carry out the position that expectation is followed the tracks of in each joint of control mechanical arm

And speed Realize the Optimal Control performance of whole system outward.

2. non-cooperative target of space robot autonomous classification according to claim 1 and catching method is characterized in that described image based on stereoscopic vision is handled and the noncooperative target relative position measurement may further comprise the steps:

(1) image filtering: respectively left and right sides camera image is carried out filtering,, obtain level and smooth left and right sides camera image to eliminate noise jamming;

(2) edge detects: two width of cloth images are carried out the edge respectively detect, obtain edge features information;

(3) straight line extracts: the image that carries out after the edge detects is carried out the straight line extraction, obtain comprising each straight line of triangular supports in each interior bar straight line information;

(4) identification of noncooperative target satellite A-frame: all the straight line information after extracting, discern six straight lines, and utilize six straight lines of the triangular supports correspondence that identifies, calculate vertex of a triangle corresponding to triangle windsurfing support;

(5) three-dimensional coupling, 3D reconstruct and pose measurement: respectively in left and right magazine 2D information, carry out the 3D reconstruct of each characteristic point according to each summit of the triangular supports that identifies, obtain the 3D coordinate of each summit of triangular supports in world coordinate system; According to the result of 3D reconstruct, further make up and arrest the object coordinate system, and calculate its position and attitude with respect to world coordinate system.

3. non-cooperative target of space robot autonomous classification according to claim 1 and catching method is characterized in that the autonomous path planning of described robot for space target acquistion may further comprise the steps:

(1) the pose deviation is calculated, and the target that stereoscopic vision is measured is with respect to the pose measurement data of pedestal, and converting into target is with respect to the pose of arm end, and judges relative pose deviation e _pAnd e _oWhether less than preset threshold ε _pAnd ε _o, if less than, then closed paw, catch target; Otherwise, other step below carrying out;

(2) prediction of target travel, according to the relative pose deviation, the motion state of real-time estimating target---next position, attitude and linear velocity, angular speed constantly;

(3) the terminal movement velocity planning of robot for space, according to the target state of being predicted, the movement velocity of planning mechanical arm end is to catch target with shortest path;

(4) robot for space is kept away unusual path planning, according to the terminal movement velocity of planning, adopts the processing method of unusual avoidance, the movement locus of each joint angle of planning mechanical arm.

4. non-cooperative target of space robot autonomous classification according to claim 1 and catching method, it is characterized in that described robot for space system coordination control makes " mechanical arm controller " and " pedestal controller " collaborative work, realizes the optimization of whole system control performance.The motion of mechanical arm control system control mechanical arm, and estimate the reaction of manipulator motion for satellite, pass to the AOCS of rail control system of satellite pedestal with the form of moment or angular momentum, by AOCS this reaction is compensated, make the attitude of satellite keep stable, the attitude measurement element is with attitude state of a control---attitude angle simultaneously, attitude angle speed, flywheel and jet state are passed to mechanical arm control system again, mechanical arm system is judged motion planning according to these states, whether plan again with decision, if attitude angle, angular speed exceeds default scope, the path that planning is described is bad, needs planning again.The control system of pedestal is also carried out feedforward compensation according to the manipulator motion that estimates to the reaction of satellite except carry out conventional FEEDBACK CONTROL according to the attitude measurement state;

5. image according to claim 2 is handled and the noncooperative target relative position measurement, it is characterized in that described image filtering adopts median filtering algorithm, described edge to detect and adopts Canny algorithm, described straight line to extract employing Hough mapping algorithm, medium filtering is for given n numerical value { a ₁, a ₂..., a _n, with they order arrangements by size.When n was odd number, that numerical value that is arranged in the centre position was called planting of this n numerical value.When n was even number, the mean value that is positioned at two numerical value in centre position was called the intermediate value of this n numerical value, and note is made med{a ₁, a ₂..., a _n, image is through behind the medium filtering, and the output of certain pixel equals the intermediate value of each pixel grey scale in this pixel field; The Canny edge detection algorithm with the Gaussian filter smoothed image, with the finite difference of single order local derviation assign to compute gradient amplitude and direction, to gradient magnitude use non-maximum suppress, with the detection of dual threshold algorithm be connected the edge; The Hough mapping algorithm with the straight line among the image space x-y, is for conversion into a bit of parameter space with the duality of point-line;

6. image according to claim 2 is handled and the noncooperative target relative position measurement, the identification that it is characterized in that described noncooperative target satellite A-frame comprises initial acquisition and two key steps of real-time tracking, and initial acquisition is by being provided with two initial reference point p to the left and right camera image that passes to ground under the telemetering channel _R1, p _R2, spaceborne program can be finished the automatic identification to the windsurfing support; After initial acquisition was set up, spaceborne program changed tracing mode immediately over to, later on can be automatically according to image, and the zone at real-time tracking windsurfing support place.Reference point p _R1For in the triangle more arbitrarily, p _R2For triangle outer and be positioned at that solar battery array lists more arbitrarily, p _R1, p _R2Directly in the remote measurement image, choose by mouse, do not have strict especially requirement;

7. image according to claim 2 is handled and the noncooperative target relative position measurement, the image coordinate that it is characterized in that the triangular supports summit that described three-dimensional coupling, 3D reconstruct and the utilization of pose measurement step identify respectively from left and right camera image, adopt least square method that three-dimensional projection model is calculated, obtain the 3D coordinate of triangular supports summit in world coordinate system, and, calculate position, the attitude of target-based coordinate system with respect to world coordinate system according to this coordinate establishing target coordinate system of 3;

8. the autonomous path planning of robot for space target acquistion according to claim 3, it is characterized in that the prediction steps of described target travel at first sets up the state equation of the motion of target, adopt next relative position, attitude and relative linear velocity, the angular speed constantly of expansion Kalman wave filter target of prediction satellite then;

9. the autonomous path planning of robot for space target acquistion according to claim 3 is characterized in that the movement velocity of the terminal movement velocity planning step of described robot for space according to following formula planning mechanical arm end:

$\left[\begin{array}{c}{E}_{V_{\mathrm{ed}}}\\ E_{{\mathrm{\ω}}_{\mathrm{ed}}}\end{array}\right]=K\left[\begin{array}{c}{E}_{r_{\mathrm{eh}}}\\ \mathrm{\Δo}\end{array}\right]+\left[\begin{array}{c}{E}_{v_h}\\ E_{{\mathrm{\ω}}_h}\end{array}\right]---\left(1\right)$

In the following formula,

Be respectively the terminal linear velocity and the angular speed of planning; K is a gain matrix, is diagonal matrix, is used to limit the terminal maximal rate of planning;

Be respectively next relative position, linear velocity and the angular speed constantly of target of prediction, the expression in the terminal coordinate system of mechanical arm, Δ O is an attitude error, is calculated as follows:

$\mathrm{\Δ}0=\frac{1}{2}(n_e\×n_h+o_e\×o_h+a_e\×a_h)=\frac{1}{2}\left[\begin{array}{c}{}^E{A}_H\left(\mathrm{2,3}\right)-{}^E{A}_H{}_{\left(\mathrm{3,2}\right)}\\ -{}^E{A}_H\left(\mathrm{1,3}\right)+{}^E{A}_H\left(\mathrm{3,1}\right)\\ {}^E{A}_H\left(\mathrm{1,2}\right)-{}^E{A}_H\left(\mathrm{2,1}\right)\end{array}\right]$

Wherein, [n _e, o _e, a _e] and [n _h, o _h, a _h] be respectively the posture changing matrix of terminal coordinate system of mechanical arm and target-based coordinate system, and

Be the handle measured by the trick camera attitude matrix with respect to terminal coordinate system.

10. the autonomous path planning of robot for space target acquistion according to claim 3 is characterized in that described robot for space keeps away unusual path planning the unusual avoidance of the dynamics of robot for space is converted into the unusual avoidance of real-time kinematics, promptly

$\left[\begin{array}{c}{E}_{V_{\mathrm{ed}}}\\ E_{{\mathrm{\ω}}_{\mathrm{ed}}}\end{array}\right]-\left({}^E\widehat{J_b}\right)\left({}^E{\mathrm{\ω}}_o\right)=\left({}^E{\hat{J}}_m\right)\·\stackrel{\·}{{\mathrm{\Θ}}_d}---\left(3\right)$

The following formula left side is the movement velocity (expression in base coordinate system) of mechanical arm end with respect to pedestal, thereby can be written as:

$\left[E,\begin{array}{c}{E}_{v_e^o}\\ {\mathrm{\ω}}_e^o\end{array}\right]={}^E{\hat{J}}_m{\stackrel{\·}{\mathrm{\Θ}}}_d---\left(4\right)$

Represent of linear velocity, angular speed the expression in base coordinate system of mechanical arm end respectively with respect to pedestal;

Be the common Jacobian matrix of mechanical arm, thus, find the solution the joint angle speed of expectation according to (4)

Images