CN105643625A - Working mechanism of single-station feeding production processing system based on mechanical arm - Google Patents

Abstract

The invention discloses a working mechanism of a single-station feeding production processing system based on a mechanical arm. The working mechanism is characterized in that the single-station feeding production processing system comprises a mechanical arm, a conveyor belt, an industrial camera, a cache library with capacity being M, a finished product warehouse with capacity being N and workpieces, wherein the mechanical arm is used for obtaining vacant amount m of the cache library and a gripper position p under a current state; and a visibility distance Vsmp is selected according to an optimal control strategy v*; and if workpiece are within the visibility distance Vsmp, unloading operation is carried out; and if the workpieces are not within the visibility distance Vsmp, service operation is carried out. The optimal control strategy v* is obtained by carrying out optimization solution on an SMDP model which is established by taking the vacant amount m of the cache library and the gripper position p as states through a strategy iteration algorithm. The working mechanism can be used for improving process balance and production efficiency of a single-station mechanical arm production line which has no fixed beat, has random material supply and has no fixed-point processing, so that basis is provided for optimal scheduling of the single-station mechanical arm production line in industrial production.

Description

A kind of working mechanism of the single site feed production and processing system of mechanically-based arm

Technical field

The present invention relates to automation field, the working mechanism of especially a kind of single site conveyer belt feed production and processing station system

Background technology

Along with the quick development of modern industry, mechanical arm automatic production line is increasingly extensive in the application of field of industrial production, as in fields such as electronic manufacture, automobile making, processing and packing, goods sortings. Particularly machine vision technique utilization on a production line, drastically increases the flexibility of system, intellectuality and automatization level. Such automatic production line generally configures one or more mechanical arm for picking or processing as concrete actuator; Configure one or more conveyer belt for conveying workpieces and packing box; Configuration industrial vision system is for different operations such as the location of workpiece, identification, dimensional measurements.

However as socioeconomic development, the particularization of production and processing, specialization, randomization to the production efficiency of automatic production line, produce versatility, intelligent and motility is had higher requirement. Traditional production line often adapts to single workpiece stream, and according to fixing beat at fixing point processing work. Adaptability and production efficiency for modes of production such as on-fixed beat, random feed, multi-varieties and small-batches are not high.

Summary of the invention

The present invention is the weak point in order to overcome prior art to exist, the working mechanism of the single site feed production and processing system of a kind of mechanically-based arm is provided, to the production efficiency of the single site mechanical arm production line processed for on-fixed beat, random feed, on-fixed point can be improved, thus providing foundation for the single site mechanical arm production line Optimized Operation in commercial production.

The present invention solves that technical problem adopts the following technical scheme that

The working mechanism of the single site feed production and processing system of a kind of mechanically-based arm of the present invention, described single site feed production and processing system includes: mechanical arm, conveyer belt, industrial camera, capacity are the buffer memory storehouse of M, capacity is warehouse for finished product and the workpiece of N;

Described mechanical arm is positioned at the side of described conveyer belt, is respectively arranged with described buffer memory storehouse and warehouse for finished product in the both sides of described mechanical arm; Described industrial camera is in the upstream of described mechanical arm, and vertical just to the workpiece on described conveyer belt, make described industrial camera vertical just to position be a P that takes pictures^cam; And described industrial camera can will be located in a P that takes pictures^camThe location of workpiece in downstream passes to described mechanical arm; The front viewpoint defining described mechanical arm is P^look, and it is positioned at described front viewpoint P^lookThe workpiece in downstream cannot be captured by described mechanical arm;

Vacant amount in definition buffer memory storehouse is m; M �� [0, M]; The position defining the handgrip of described mechanical arm is p; As p=1, represent that the position of handgrip of described mechanical arm is in warehouse for finished product; As p=0, represent that the position p of handgrip of described mechanical arm is in buffer memory storehouse; The united state S of described system of processing it is made up of the position p of vacant amount m and handgrip_m,p; Definition is with described front viewpoint P^lookOne section of observed range for starting point is forward sight distanceDefinition workpiece is through described forward sight distanceThe time spent is the forward sight time

Defining described mechanical arm working range on described conveyer belt is W^pick, and at described working range W^pickInside carry out unloading operation, buffer memory storehouse carries out service operations; Described unloading operation is from capturing described conveyer belt to described buffer memory storehouse by described workpiece; Described service operations is to put in described warehouse for finished product after being processed by the workpiece in described buffer memory storehouse; The time having defined unloading operation is discharge timeThe time having defined service operations is service timeAfter completing unloading operation, the handgrip of described mechanical arm is in described buffer memory storehouse; After completing service operations, the handgrip of described mechanical arm is in described warehouse for finished product; Being characterized in, described working mechanism carries out as follows:

Step 1, described system of processing bring into operation, and workpiece arrives mechanical arm operation interval at random; Set line speed, mechanical arm speed, workpiece arrival rate, work pieces process rate; Defined variable i, and initialize i=1;

Step 2, described mechanical arm obtain the vacant amount m of i & lt_iWith handgrip position p_i, according to optimal control policy v^*Select i & lt forward sight distance

Step 3, judge the optimum forward sight distance of described i & ltInside whether there is workpiece, if there being workpiece, then the described mechanical arm optimum forward sight distance to being in described i & ltInterior near front viewpoint P^lookWorkpiece carry out unloading operation after, perform step 4; If there is no workpiece, then after the workpiece in described buffer memory storehouse is carried out service operations by described mechanical arm, perform step 5;

Step 4, judge near front viewpoint P^lookThe i & lt discharge time of workpieceWhether more than described near front viewpoint P^lookI & lt forward sight time of workpieceIf more than, then after i+1 being assigned to i, return step 2 and perform; Otherwise, through the unloading time delay of i & ltAfter, i+1 is assigned to i, and returns step 2 and perform;

Step 5, judge near front viewpoint P^lookI & lt service time of workpieceWhether more than described near front viewpoint P^lookI & lt forward sight time of workpieceIf more than, then after i+1 being assigned to i, return step 2 and perform; Otherwise, time delay is serviced through i & ltAfter, i+1 is assigned to i, and returns step 2 and perform.

The feature of the working mechanism of the single site feed production and processing system of mechanically-based arm of the present invention lies also in:

Described system of processing is as state to control described forward sight distance using described vacant amount m and handgrip (7) position pLength, and then control the mode of operation of described mechanical arm (1);

When vacant amount m is less, then described mechanical arm (1) chooses shorter forward sight distanceDescribed mechanical arm trends towards service operations;

When vacant amount m is bigger, then described mechanical arm (3) chooses longer forward sight distanceDescribed mechanical arm trends towards unloading operation.

Described optimal control policy v^*Obtain according to the following steps:

Step 1, described system of processing are using the vacant amount in buffer memory storehouse and handgrip position as state, forward sight distanceChoose as action, set up half Markov decision model;

Step 2, it is optimized by double Markov model of Policy iteration algorithm and solves, obtain optimal control policy v^*��

Described half Markov model is set up according to the following steps:

Step 1, define the state in described buffer memory storehouse for its vacant amount m, then the state space in described buffer memory storehouse is ��₁, and have ��₁=0,1 ..., M};

The state space of definition handgrip position p is ��₂, and have ��₂={ 0,1}; As p=0, represent that the position p of handgrip of described mechanical arm is in buffer memory storehouse, as p=1, represent that the position of handgrip of described mechanical arm is in warehouse for finished product;

Define the united state S of described system of processing_m,pState space be ��, and have ��=��₁����₂;

Step 2, the former apparent distanceAs described system of processing at united state S_m,pUnder action, a steady control strategy v of described system of processing is all of united state to the mapping of action, and hasWhereinRepresent when the vacant amount in buffer memory storehouse (4) is m, the action of system of processing when the position of handgrip (7) is in buffer memory storehouse (4);Represent when the vacant amount in buffer memory storehouse (4) is m, the action of system of processing when the position of handgrip (7) is in warehouse for finished product (5); Definition All Policies v set be ��, ��=v | v=(v (1), v (2), v (3) ..., v (N)), v (i) �� D}; Described system of processing is absent from S_M,0And S_0,1The two state, then corresponding action is designated as NaN;

Step 3, definition T_iFor the i & lt decision-making moment of described system of processing, make T₀=0; Defining unloading operation time delay when described system of processing carries out the process of i & lt Markov markov property isService operations time delay is

When the i & lt of described mechanical arm operates as unloading operation, then i+1 time decision-making moment T_i+1For:

$T_{i+1}=T_i+max\{t_{S_{m_i,p_i}},t_{m_i,p_i}^{load}\}---\left(1\right)$

I+1 time united state is X_i+1=S_m-1,0, representing that handgrip (7) position is p=0, the vacant amount in buffer memory storehouse (4) is m-1;

I & lt unloading operation time delay is:

When the i & lt of described mechanical arm operates as service operations, then i+1 time decision-making moment T_i+1For:

$T_{i+1}=T_i+max\{t_{S_{m_i,p_i}},t_{m_i,p_i}^{ser}\}---\left(3\right)$

I+1 time united state is X_i+1=S_m+1,0; Representing that handgrip (7) position is p=0, the vacant amount in buffer memory storehouse (4) is m+1;

I & lt service operations time delay is:

Define half Markov core Q^v(t) be:

$Q^v\left(t\right)={\[Q(s_{m,p},v_{s_{m,p}},s_{m^{\′},p^{\′}},t)\]}_{s_{m,p},s_{m^{\′},p^{\′}}\∈\mathrm{\Φ}}=P^v\⊗F^v\left(t\right)---\left(5\right)$

In formula (5),The operator that representing matrix is multiplied; T represents system time; P^vRepresent Markov embedded chain transfer matrix under strategy v, F^vT () represents the distribution of residence-time matrix under strategy v, and have:

$P^v={\[p_{s_{m,p}{s}_{m^{\′},p^{\′}}}\left(v_{s_{m,p}}\right)\]}_{s_{m,p}{s}_{m^{\′},p^{\′}}\∈\mathrm{\Φ}}---\left(6\right)$

$F^v\left(t\right)={\[F_{s_{m,p}{s}_{m^{\′},p^{\′}}}(t,v_{s_{m,p}})\]}_{s_{m,p}{s}_{m^{\′},p^{\′}}\∈\mathrm{\Φ}}---\left(7\right)$

In formula (6)It is from state S_m,pTake actionTransfer to S_m',p'Probability, in formula (7)It is at state S_m,pTake actionIt is S to NextState_m',p'Distribution of residence-time;

Step 4, definition expected performance function f^vFor:

$f^v={\[f(s_{m,p},v_{s_{m,p}},s_{m^{\′},p^{\′}})\]}_{s_{m,p},s_{m^{\′},p^{\′}}\∈\mathrm{\Φ}}---\left(8\right)$

In formula (8),Represent that system of processing is taken actionFrom united state S_m,pTransfer to NextState S_m',p'Unit interval expectation cost;

Step 5, foundation SMDP model as shown in formula (9):

X=(X_t,��,D,Q^v(t),f^v)(9)

In formula (9), X_tRepresent the state procedure of t system of processing.

Described Policy iteration algorithm sequentially includes the following steps:

Step 1, according to half Markov core Q^v(t) and expected performance function f^v, define Equivalent InfinitesimalWith performance matrix of equal valueAnd by SMDP model X=(X_t,��,D,Q^v(t),f^v) be converted to continuous time Markov decision making process�� is discount factor, �� �� [0,1];

Step 2, defined variable k, define v_kKth time strategy for system of processing; Initialize k=0; And the 0th tactful v₀For:

Step 3, formula (11) is utilized to obtain kth time strategy v_kPerformance potential vector

$(\mathrm{\α}I-A_{\mathrm{\α}}^{v_k}+{\mathrm{\ρ}}_{\mathrm{\α}}{\mathrm{e\π}}_{\mathrm{\α}}^{v_k})g_{\mathrm{\α}}^{v_k}=f_{\mathrm{\α}}^{v_k}---\left(11\right)$

In formula (11); I is unit diagonal matrix; E is unit column vector; And have:

$h_{\mathrm{\α}}^{v_k}=(I-Q_{\mathrm{\α}}^{v_k})e/\mathrm{\α}---\left(12\right)$

${\mathrm{\ρ}}_{\mathrm{\α}}=max\{h_{\mathrm{\α}}^{v_k}{\left(i\right)}^{-1},i=1,2...N,v_k\∈\mathrm{\Ω},\mathrm{\α}\∈\[0,1\]\}---\left(13\right)$

Represent the average waiting time at state i,Represent the steady-state distribution of kth time in Markov decision making process of equal value; And have:

$\begin{array}{c}{A}_{\mathrm{\α}}^{v_k}{\mathrm{\π}}_{\mathrm{\α}}^{v_k}=0\\ {\mathrm{\π}}_{\mathrm{\α}}^{v_k}e=1\\ A_{\mathrm{\α}}^{v_k}e=0\end{array}---\left(14\right)$

Step 4, utilize formula (15) obtain+1 tactful v of kth_k+1:

$v_{k+1}\∈\arg \underset{v\∈\mathrm{\Ω}}{min}\{A_{\mathrm{\α}}^v{g}_{\mathrm{\α}}^{v_k}+f_{\mathrm{\α}}^v\}---\left(15\right)$

Step 5, utilize formula (11) obtain+1 tactful v of kth_k+1Performance potential vector

Step 6, judgementWhether set up, �� is a less constant; If setting up, then it represents that+1 tactful v of the kth obtained_k+1It is optimal control policy v^*, and exit algorithm, otherwise; After k+1 is assigned to k, return step 4.

Compared with the prior art, beneficial effects of the present invention is embodied in:

The present invention adopts a kind of single site mechanical arm production and processing system working mechanism based on half Markov model, can effectively process on-fixed beat, on-fixed picks point, workpiece arrives at random production line processing problems, improve balance and the production efficiency of each operation of production line.

1, the present invention adopts half Markov model that single site mechanical arm production and processing system is modeled, the random workpiece stream arrived is divided into the independent workpiece sequence meeting Markov process of a lot of segment, comparing traditional method according to fixing beat, workpiece to be processed, the present invention more effective process can arrive the picking of workpiece, processing problems at random.

2, before the present invention apparent distance control the action as system of processing, it is possible to the situation according to the real-time workpiece of production line, buffer memory storehouse and warehouse for finished product, adjust Processing Strategies. It is effectively increased the balance between each operation of production line, strengthens the robustness of production line.

3, the present invention adopts Policy iteration algorithm that the SMDP model of system of processing is optimized to solve, compare other algorithms, and Policy iteration algorithm avoids local optimum, fast convergence rate, it is possible to reduces system of processing and solves the time, improves production efficiency.

4, mechanical arm of the present invention adopts may move and picks mode, calculates according to the location of workpiece on operation interval and picks collision point, it is possible to reduces uninstall process and waits the time of workpiece.

Accompanying drawing explanation

Fig. 1 is single site feed production and processing system schematic of the present invention;

Fig. 2 is working mechanism's flow chart of single site feed production and processing system of the present invention.

Detailed description of the invention

Referring to Fig. 1, in the present embodiment, single site feed production and processing system includes: mechanical arm 1, conveyer belt 2, industrial camera 3, capacity are the buffer memory storehouse 4 of M, capacity is warehouse for finished product 5 and the workpiece 6 of N;

Mechanical arm 1 is positioned at the side of conveyer belt 2, and mechanical arm is 6DOF cascade machine mechanical arm, equipped with pneumatic gripping device 7 on the ring flange of mechanical arm 1. Definition mechanical arm speed is V_robot, in the implementation case, V_robot=20cm/s. The buffer memory storehouse 4 that capacity is M and the warehouse for finished product 5 that capacity is N it is respectively arranged with in the both sides of mechanical arm 1; In the implementation case, M=4, N=2147483648. Industrial camera 3 is in the upstream of mechanical arm 1, and vertical just to the workpiece 6 on conveyer belt 2; Conveyer belt is by driven by servomotor, and speed can accurately regulate. Definition conveyer belt transporting velocity is V_con, in the implementation case, V_con=10cm/s; Definition mechanical arm 1 working range on conveyor belt 2 is W^pick, wherein W^pickLeft end point isRight endpoint isTaking pictures a little for P of definition industrial camera 3^cam;It is positioned at a P that takes pictures^camThe information such as the location of workpiece in downstream can pass to mechanical arm 1; The front viewpoint defining described mechanical arm (1) is P^look, and it is positioned at front viewpoint downstream P^lookWorkpiece will lose;

Vacant amount in definition buffer memory storehouse 4 is m, m �� [0,4]; The position of the handgrip 7 of definition mechanical arm 1 is p; As p=1, represent that the position of handgrip 7 of mechanical arm 1 is in warehouse for finished product 5; As p=0, represent that the position p of handgrip 7 of mechanical arm 1 is in buffer memory storehouse 4; Definition mechanical arm handgrip coordinate is P_p; The united state S of system of processing it is made up of the position p of vacant amount m and handgrip 7_m,p; Viewpoint P before definition^lookOne section of observed range for starting point is forward sight distanceD=[0, l_max]; Definition workpiece 6 is through forward sight distanceThe time spent is the forward sight timeAnd have:

$P^{look}=P_{right}^{pick}+\frac{\left|\right|P_p-P_{right}^{pick}\left|\right|}{V_{robot}}{V}_{con}---\left(1\right)$

l_max=P^cam-P^look(2)

$t_{S_{m,p}}=\frac{v_{S_{m,p}}}{V_{con}}---\left(3\right)$

Mechanical arm is at working range W^pickInside carry out unloading operation, buffer memory storehouse 4 carries out service operations; Unloading operation is from capturing conveyer belt 2 to buffer memory storehouse 4 by workpiece 6; Service operations is to put in warehouse for finished product 5 after being processed by the workpiece 6 in buffer memory storehouse 4; The time having defined unloading operation is discharge timeThe time having defined service operations is service timeAfter completing unloading operation, the handgrip 7 of mechanical arm 1 is in buffer memory storehouse 4; After completing service operations, the handgrip 7 of mechanical arm 1 is in warehouse for finished product 5;

System of processing controls forward sight distance using vacant amount m and handgrip 7 position p as stateLength and the mode of operation of mechanical arm 1; When vacant amount m is less, it was shown that the storage pressure in buffer memory storehouse 4 is less, then mechanical arm 1 chooses shorter forward sight distanceMechanical arm 1 trends towards service operations, and the discharge pressure to reduce conveyer belt 2 improves the utilization rate in buffer memory storehouse 4; When vacant amount m is bigger, it was shown that the storage pressure in buffer memory storehouse 4 is relatively big, then mechanical arm 1 chooses longer forward sight distanceMechanical arm 1 trends towards unloading operation, to alleviate the storage pressure in buffer memory storehouse 4.

Described system of processing is using the vacant amount in buffer memory storehouse 4 and handgrip position 7 as state, forward sight distanceChoose as action, set up half Markov decision model according to the following steps:

Step 1, definition buffer memory storehouse 4 state be its vacant amount m, then the state space in buffer memory storehouse 4 is ��₁, and have ��₁=0,1 ..., M};

The state space of definition handgrip 7 position p is ��₂, and have ��₂=0,1}, as p=1, represent that the position of handgrip 7 of mechanical arm 1 is in warehouse for finished product 5; As p=0, represent that the position p of handgrip 7 of mechanical arm 1 is in buffer memory storehouse 4;

The united state S of definition system of processing_m,pState space be ��, and have ��=��₁����₂;

Step 2, the former apparent distanceAs system of processing at united state S_m,pUnder action, a steady control strategy v of described system of processing is all of united state to the mapping of action, and has:

In formula (4)Represent when vacant amount in buffer memory storehouse 4 is m, the action of system of processing during the position p=0 of handgrip 7;Represent when vacant amount in buffer memory storehouse 4 is m, the action of system of processing during the position p=1 of handgrip 7; The set of definition All Policies v is ��; ��=v | v=(v (1), v (2), v (3) ..., v (N)), v (i) �� D}. System of processing is absent from S_M,0And S_0,1The two state, then corresponding action is designated as NaN;

Step 3, definition T_iMoment, wherein T is shifted for described system of processing i & lt₀=0. Defining unloading operation time delay when described system of processing carries out the process of Markov markov property isService operations time delay is

When the operation of described mechanical arm i & lt is for unloading operation, then next decision-making moment T_i+1For:

$T_{i+1}=T_i+max\{t_{S_{m_i,p_i}},t_{m_i,p_i}^{load}\}---\left(5\right)$

NextState is X_i+1=S_m-1,0;System uninstallation delay time is:

When the operation of described mechanical arm i & lt is for service operations, then next decision-making moment T_i+1For:

$T_{i+1}=T_i+max\{t_{S_{m_i,p_i}},t_{m_i,p_i}^{ser}\}---\left(7\right)$

NextState is X_i+1=S_m-1,0; System service delay time is:

Define half Markov core Q^v(t) be:

$Q^v\left(t\right)={\[Q(s_{m,p},v_{s_{m,p}},s_{m^{\′},p^{\′}},t)\]}_{s_{m,p},s_{m^{\′},p^{\′}}\∈\mathrm{\Φ}}=P^v\⊗F^v\left(t\right)---\left(9\right)$

In formula (9),The operator that representing matrix is multiplied; P^vRepresent Markov embedded chain transfer matrix under strategy v, F^vT () represents the distribution of residence-time matrix under strategy v, and have:

$P^v={\[p_{s_{m,p}{s}_{m^{\′},p^{\′}}}\left(v_{s_{m,p}}\right)\]}_{s_{m,p}{s}_{m^{\′},p^{\′}}\∈\mathrm{\Φ}}---\left(10\right)$

$F^v\left(t\right)={\[F_{s_{m,p}{s}_{m^{\′},p^{\′}}}(t,v_{s_{m,p}})\]}_{s_{m,p}{s}_{m^{\′},p^{\′}}\∈\mathrm{\Φ}}---\left(11\right)$

In formula (10)It is from state S_m,pTake actionTransfer to S_m',p'Probability, and have:

$\begin{array}{c}{p}_{S_{0,0}{S}_{1,1}}\left(v_{S_{0,0}}\right)=p_{S_{M,1}{S}_{M-1,0}}\left(v_{S_{M,1}}\right)\≡1\\ p_{S_{m,p}{S}_{m-1,0}}\left(v_{S_{m,p}}\right)=1-e^{-{\mathrm{\λv}}_{S_{m,p}}},m=1,2,\mathrm{...},M-1\\ p_{S_{m,p}{S}_{m+1,1}}\left(v_{S_{m,p}}\right)=e^{-{\mathrm{\λv}}_{S_{m,p}}},m=1,2,\mathrm{...},M-1\end{array}---\left(12\right)$

In formula (11)It is at state S_m,pTake actionIt is S to NextState_m',p'Distribution of residence-time.

Step 4, definition expected performance function f^vFor:

$f^v={\[f(s_{m,p},v_{s_{m,p}},s_{m^{\′},p^{\′}})\]}_{s_{m,p},s_{m^{\′},p^{\′}}\∈\mathrm{\Φ}}---\left(13\right)$

In formula (13),Represent that system of processing is taken actionFrom united state S_m,pTransfer to NextState S_m',p'Unit interval expectation cost;

Step 5, foundation SMDP model as shown in Equation 14:

X=(X_t,��,D,Q^v(t),f^v)(14)

In formula (14), X_tRepresent the state procedure of t.

According to following steps described half Markov model is optimized by Policy iteration algorithm and solves, obtain optimal control policy v^*, and by optimal control policy v^*It is cured in described mechanical arm 1.

Step 1, according to half Markov core Q^v(t) and expected performance function f^v, define Equivalent InfinitesimalWith performance matrix of equal valueAnd by SMDP model X=(X_t,��,D,Q^v(t),f^v) be converted to continuous time Markov decision making processWherein �� is discount factor, �� �� [0,1].

Step 2, defined variable k, define v_kKth time strategy for system of processing; Initialize k=0; And the 0th tactful v₀For:

Step 3, formula (16) is utilized to obtain kth time strategy v_kPerformance potential vector

$(\mathrm{\α}I-A_{\mathrm{\α}}^{v_k}+{\mathrm{\ρ}}_{\mathrm{\α}}{\mathrm{e\π}}_{\mathrm{\α}}^{v_k})g_{\mathrm{\α}}^{v_k}=f_{\mathrm{\α}}^{v_k}---\left(16\right)$

In formula (16); I is unit diagonal matrix; E is unit column vector; And have:

$h_{\mathrm{\α}}^{v_k}=(I-Q_{\mathrm{\α}}^{v_k})e/\mathrm{\α}---\left(17\right)$

${\mathrm{\ρ}}_{\mathrm{\α}}=max\{h_{\mathrm{\α}}^{v_k}{\left(i\right)}^{-1},i=1,2...N,v_k\∈\mathrm{\Ω},\mathrm{\α}\∈\[0,1\]\}---\left(18\right)$

Represent the average waiting time at state i,Represent the steady-state distribution of kth time in Markov decision making process of equal value; And have:

$\begin{array}{c}{A}_{\mathrm{\α}}^{v_k}{\mathrm{\π}}_{\mathrm{\α}}^{v_k}=0\\ {\mathrm{\π}}_{\mathrm{\α}}^{v_k}e=1\\ A_{\mathrm{\α}}^{v_k}e=0\end{array}---\left(19\right)$

Step 4, utilize formula (20) obtain+1 tactful v of kth_k+1:

$v_{k+1}\∈\arg \underset{v\∈\mathrm{\Ω}}{min}\{A_{\mathrm{\α}}^v{g}_{\mathrm{\α}}^{v_k}+f_{\mathrm{\α}}^v\}---\left(20\right)$

Step 5, utilize formula (16) obtain+1 tactful v of kth_k+1Performance potential vector

Step 6, judgementWhether set up, �� is a less constant; If setting up, then it represents that+1 tactful v of the kth obtained_k+1It is optimal control policy v^*, and exit algorithm, otherwise; After k+1 is assigned to k, return step 4.

By obtained optimal control policy v^*After being cured in described mechanical arm 1, the working mechanism of system of processing carries out as follows:

Step 1, described system of processing bring into operation, and set line speed V_con=10cm/s, mechanical arm speed V_robot=20cm/s, workpiece arrival rate ��=1, work pieces process rate ��=2; Defined variable i, and initialize i=1.

Step 2, described mechanical arm 1 obtain the vacant amount m of i & lt_iWith handgrip 7 position p_i, according to optimal control policy v^*Select i & lt forward sight distance

Step 3, judge the optimum forward sight distance of described i & ltInside whether there is workpiece, if there being workpiece, then the described mechanical arm 1 forward sight distance to being in described i & ltInterior near front viewpoint P^lookWorkpiece carry out unloading operation after, perform step 4; If there is no workpiece, then after the workpiece in described buffer memory storehouse 4 is carried out service operations by described mechanical arm 1; Perform step 5;

Wherein uninstall process is divided into and picks, places and wait three processes; Select in forward sight distanceFrom front viewpoint P^lookA nearest workpiece is as target workpiece, and definition observes that target workpiece coordinate is P^work, handgrip is from initial position P_pSetting out, following the tracks of the collision point coordinate definition picking workpiece isDefinition is mobile picks the timeMovement is standing timeWith the waiting time it isDefinition target workpiece arrives front viewpoint P^lookTime be the time of adventThen:

DefinitionFor the separation of the workpiece time of advent, and have:

$T_p^{able}=\frac{\left|\right|P_p-P_{left}^{pick}\left|\right|}{V_{robot}}+\frac{P_{left}^{pick}-P^{look}}{V_{con}}---\left(22\right)$

IfThen directly pick workpiece;Then collision pointThe time is picked with mobileMeet equation group:

$\left\{\begin{array}{c}\left|\right|P^{work}-P_p^{pick}\left|\right|=t_p^{pick}{V}_{con}\\ \left|\right|P_p-P_p^{pick}\left|\right|=t_p^{pick}{V}_{robot}\end{array}\right.---\left(23\right)$

Waiting time is:

$t_p^{wait}=0---\left(24\right)$

IfThen mechanical arm handgrip first moves to operation interval left end pointWait that workpiece arrives to pick, then the waiting time is again:

Mobile pick the time and be

$t_p^{pick}=\frac{\left|\right|P_{left}^{pick}-P_p\left|\right|}{V_{robot}}---\left(26\right)$

Handgrip goes back to buffer memory storehouse 4 after picking workpiece and places, and movement is standing time:

$t_p^{place}=\left\{\begin{array}{cc}{t}_0^{pick}& p=0\\ \frac{\left|\right|P_1^{pick}-P_0\left|\right|}{V_{robot}}& p=1\end{array}\right.---\left(27\right)$

So the time of i & lt unloading operationFor:

$t_{m_i,p_i}^{load}=t_{p_i}^{pick}+t_{p_i}^{place}+t_{p_i}^{wait}---\left(28\right)$

Then the next decision-making moment is:

NextState is X_i+1=S_m-1,0; The system delay time is:

Service operations is divided into mobile taking-up workpiece and the course of processing, and being divided into initial position equally is buffer memory storehouse 4 and 5 two kinds of situations of warehouse for finished product; If p=0, namely initial position is buffer memory storehouse 4, then mechanical arm is directly postponed and taken out a workpiece in warehousing 4 and be processed, and is positioned over by workpiece in warehouse for finished product 5 after machining; If p=1, namely initial position is warehouse for finished product 5, then mechanical arm first moves to buffer memory storehouse 4 and takes out workpiece and be processed; The time of definition i & lt service operations isWherein obeying fixing random distribution process time, its time isDefinition is T from warehouse for finished product to buffer memory storehouse spended time^bank_buffer, then

${\mathrm{\τ}}_{m_i,p_i}^{ser}=\left\{\begin{array}{cc}{t}_i^{process}& p=0\\ t_i^{process}+T^{bank\_buffer}& p=1\end{array}\right.---\left(31\right)$

Then the next decision-making moment is:

$T_{i+1}=T_i+max\{t_{S_{m_i,p_i}},{\mathrm{\τ}}_{m_i,p_i}^{ser}\}---\left(32\right)$

Then next state is X_i+1=S_m+1,1, the system delay time is:

Step 4, judge near front viewpoint P^lookThe i & lt discharge time of workpieceWhether more than described near front viewpoint P^lookI & lt forward sight time of workpieceIf more than, then after i+1 being assigned to i, return step 2 and perform; Otherwise, through the unloading time delay of i & ltAfter, i+1 is assigned to i, and returns step 2 and perform;

Step 5, judge near front viewpoint P^lookI & lt service time of workpieceWhether more than described near front viewpoint P^lookI & lt forward sight time of workpieceIf more than, then after i+1 being assigned to i, return step 2 and perform; Otherwise, time delay is serviced through i & ltAfter, i+1 is assigned to i, and returns step 2 and perform.

Claims (5)

1. a working mechanism for the single site feed production and processing system of mechanically-based arm, described single site feed production and processing system includes: mechanical arm (1), conveyer belt (2), industrial camera (3), capacity are the buffer memory storehouse (4) of M, capacity is warehouse for finished product (5) and the workpiece (6) of N;

Described mechanical arm (1) is positioned at the side of described conveyer belt (2), is respectively arranged with described buffer memory storehouse (4) and warehouse for finished product (5) in the both sides of described mechanical arm (1); Described industrial camera (3) is in the upstream of described mechanical arm (1), and vertical just to the workpiece (6) on described conveyer belt (2), make described industrial camera (3) vertical just to position be a P that takes pictures^cam; And described industrial camera (3) can will be located in a P that takes pictures^camThe location of workpiece in downstream passes to described mechanical arm (1); The front viewpoint defining described mechanical arm (1) is P^look, and it is positioned at described front viewpoint P^lookThe workpiece in downstream cannot be captured by described mechanical arm (1);

Vacant amount in definition buffer memory storehouse (4) is m; M �� [0, M]; The position of the handgrip (7) defining described mechanical arm (1) is p; As p=1, represent that the position of handgrip (7) of described mechanical arm (1) is in warehouse for finished product (5); As p=0, represent that the position p of handgrip (7) of described mechanical arm (1) is in buffer memory storehouse (4); The united state S of described system of processing it is made up of the position p of vacant amount m and handgrip (7)_m,p; Definition is with described front viewpoint P^lookOne section of observed range for starting point is forward sight distance Definition workpiece (6) is through described forward sight distanceThe time spent is the forward sight time

Defining the described mechanical arm (1) working range on described conveyer belt (2) is W^pick, and at described working range W^pickInside carry out unloading operation, buffer memory storehouse (4) carry out service operations;Described unloading operation is from capturing described conveyer belt (2) to described buffer memory storehouse (4) by described workpiece (6); Described service operations is to put in described warehouse for finished product (5) after being processed by the workpiece (6) in described buffer memory storehouse (4); The time having defined unloading operation is discharge timeThe time having defined service operations is service timeAfter completing unloading operation, the handgrip (7) of described mechanical arm (1) is in described buffer memory storehouse (4); After completing service operations, the handgrip (7) of described mechanical arm (1) is in described warehouse for finished product (5);

It is characterized in that, described working mechanism carries out as follows:

Step 1, described system of processing bring into operation, and workpiece arrives mechanical arm operation interval at random; Set line speed, mechanical arm speed, workpiece arrival rate, work pieces process rate; Defined variable i, and initialize i=1;

Step 2, described mechanical arm (1) obtain the vacant amount m of i & lt_iWith handgrip (7) position p_i, according to optimal control policy v^*Select i & lt forward sight distance

Step 3, judge the optimum forward sight distance of described i & ltInside whether there is workpiece, if there being workpiece, then the described mechanical arm (1) the optimum forward sight distance to being in described i & ltInterior near front viewpoint P^lookWorkpiece carry out unloading operation after, perform step 4; If there is no workpiece, then after the workpiece in described buffer memory storehouse (4) is carried out service operations by described mechanical arm (1), perform step 5;

Step 4, judge near front viewpoint P^lookThe i & lt discharge time of workpieceWhether more than described near front viewpoint P^lookI & lt forward sight time of workpieceIf more than, then after i+1 being assigned to i, return step 2 and perform; Otherwise, through the unloading time delay of i & ltAfter, i+1 is assigned to i, and returns step 2 and perform;

Step 5, judge near front viewpoint P^lookI & lt service time of workpieceWhether more than described near front viewpoint P^lookI & lt forward sight time of workpieceIf more than, then after i+1 being assigned to i, return step 2 and perform; Otherwise, time delay is serviced through i & ltAfter, i+1 is assigned to i, and returns step 2 and perform.

2. the working mechanism of the single site feed production and processing system of mechanically-based arm according to claim 1, is characterized in that,

Described system of processing is as state to control described forward sight distance using described vacant amount m and handgrip (7) position pLength, and then control the mode of operation of described mechanical arm (1);

When vacant amount m is less, then described mechanical arm (1) chooses shorter forward sight distanceDescribed mechanical arm (1) trends towards service operations;

When vacant amount m is bigger, then described mechanical arm (3) chooses longer forward sight distanceDescribed mechanical arm (1) trends towards unloading operation.

3. the working mechanism of the single site feed production and processing system of mechanically-based arm according to claim 1, is characterized in that,

Described optimal control policy v^*Obtain according to the following steps:

Step 1, described system of processing are using the vacant amount of buffer memory storehouse (4) and handgrip (7) position as state, forward sight distanceChoose as action, set up half Markov decision model;

Step 2, it is optimized by double Markov model of Policy iteration algorithm and solves, obtain optimal control policy v^*��

4. the working mechanism of the single site feed production and processing system of mechanically-based arm according to claim 3, is characterized in that:

Described half Markov model is set up according to the following steps:

Step 1, define the state of described buffer memory storehouse (4) for its vacant amount m, then the state space of described buffer memory storehouse (4) is ��₁, and have ��₁=0,1 ..., M};

The state space of definition handgrip (7) position p is ��₂, and have ��₂={ 0,1}; As p=0, represent that the position p of handgrip (7) of described mechanical arm (1) is in buffer memory storehouse (4), as p=1, represent that the position of handgrip (7) of described mechanical arm (1) is in warehouse for finished product (5);

Define the united state S of described system of processing_m,pState space be ��, and have ��=��₁����₂;

Step 2, the former apparent distanceAs described system of processing at united state S_m,pUnder action, a steady control strategy v of described system of processing is all of united state to the mapping of action, and hasWherein Represent when the vacant amount in buffer memory storehouse (4) is m, the action of system of processing when the position of handgrip (7) is in buffer memory storehouse (4);Represent when the vacant amount in buffer memory storehouse (4) is m, the action of system of processing when the position of handgrip (7) is in warehouse for finished product (5); Definition All Policies v set be ��, ��=v | v=(v (1), v (2), v (3) ..., v (N)), v (i) �� D}; Described system of processing is absent from S_M,0And S_0,1The two state, then corresponding action is designated as NaN;

Step 3, definition T_iFor the i & lt decision-making moment of described system of processing, make T₀=0; Defining unloading operation time delay when described system of processing carries out the process of i & lt Markov markov property isService operations time delay is

When the i & lt of described mechanical arm (1) operates as unloading operation, then i+1 time decision-making moment T_i+1For:

$T_{i+1}=T_i+max\{t_{S_{m_i,p_i}},t_{m_i,p_i}^{load}\}---\left(1\right)$

I+1 time united state is X_i+1=S_m-1,0, representing that handgrip (7) position is p=0, the vacant amount in buffer memory storehouse (4) is m-1;

I & lt unloading operation time delay is:

When the i & lt of described mechanical arm (1) operates as service operations, then i+1 time decision-making moment T_i+1For:

$T_{i+1}=T_i+max\{t_{S_{m_i,p_i}},t_{m_i,p_i}^{ser}\}---\left(3\right)$

I+1 time united state is X_i+1=S_m+1,0; Representing that handgrip (7) position is p=0, the vacant amount in buffer memory storehouse (4) is m+1;

I & lt service operations time delay is:

Define half Markov core Q^v(t) be:

$Q^v\left(t\right)={\[Q(s_{m,p},v_{s_{m,p}},s_{m^{\′},p^{\′}},t)\]}_{s_{m,p},s_{m^{\′},p^{\′}}\∈\mathrm{\Φ}}=P^v\⊗F^v\left(t\right)---\left(5\right)$

In formula (5),The operator that representing matrix is multiplied; T represents system time; P^vRepresent Markov embedded chain transfer matrix under strategy v, F^vT () represents the distribution of residence-time matrix under strategy v, and have:

$P^v={\[p_{s_{m,p}{s}_{m^{\′},p^{\′}}}\·\left(v_{s_{m,p}}\right)\]}_{s_{m,p}{s}_{m^{\′},p^{\′}}\∈\mathrm{\Φ}}---\left(6\right)$

$F^v\left(t\right)={\[F_{s_{m,p}{s}_{m^{\′},p^{\′}}}(t,v_{s_{m,p}})\]}_{s_{m,p}{s}_{m^{\′},p^{\′}}\∈\mathrm{\Φ}}---\left(7\right)$

In formula (6)It is from state S_m,pTake actionTransfer to S_m',p'Probability, in formula (7)It is at state S_m,pTake actionIt is S to NextState_m',p'Distribution of residence-time;

Step 4, definition expected performance function f^vFor:

$f^v={\[f(s_{m,p},v_{s_{m,p}},s_{m^{\′},p^{\′}})\]}_{s_{m,p},s_{m^{\′},p^{\′}}\∈\mathrm{\Φ}}---\left(8\right)$

In formula (8),Represent that system of processing is taken actionFrom united state S_m,pTransfer to NextState S_m',p'Unit interval expectation cost;

Step 5, foundation SMDP model as shown in formula (9):

X=(X_t,��,D,Q^v(t),f^v)(9)

In formula (9), X_tRepresent the state procedure of t system of processing.

5. the working mechanism of the single site feed production and processing system of mechanically-based arm according to claim 3, is characterized in that:

Described Policy iteration algorithm sequentially includes the following steps:

Step 1, according to half Markov core Q^v(t) and expected performance function f^v, define Equivalent InfinitesimalWith performance matrix of equal valueAnd by SMDP model X=(X_t,��,D,Q^v(t),f^v) be converted to continuous time Markov decision making process�� is discount factor, �� �� [0,1];

Step 2, defined variable k, define v_kKth time strategy for system of processing; Initialize k=0; And the 0th tactful v₀For:

$v_0=\left[\begin{array}{cc}{l}_{\max }& NaN\\ l_{\max }& l_{\max }\\ \·& l_{\max }\\ \·& \·\\ \·& \·\\ l_{\max }& \·\\ \·& l_{\max }\\ \·& \·\\ \·& \·\\ l_{\max }& \·\\ NaN& l_{\max }\end{array}\right]---\left(10\right)$

Step 3, formula (11) is utilized to obtain kth time strategy v_kPerformance potential vector

$(\mathrm{\α}I-A_{\mathrm{\α}}^{v_k}+{\mathrm{\ρ}}_{\mathrm{\α}}{\mathrm{e\π}}_{\mathrm{\α}}^{v_k})g_{\mathrm{\α}}^{v_k}=f_{\mathrm{\α}}^{v_k}---\left(11\right)$

In formula (11); I is unit diagonal matrix; E is unit column vector; And have:

$h_{\mathrm{\α}}^{v_k}=(I-Q_{\mathrm{\α}}^{v_k})e/\mathrm{\α}---\left(12\right)$

${\mathrm{\ρ}}_{\mathrm{\α}}=max\{h_{\mathrm{\α}}^{v_k}{\left(i\right)}^{-1},i=1,2...N,v_k\∈\mathrm{\Ω},\mathrm{\α}\∈\[0,1\]\}---\left(13\right)$

Represent the average waiting time at state i,Represent the steady-state distribution of kth time in Markov decision making process of equal value;And have:

$\begin{array}{c}{A}_{\mathrm{\α}}^{v_k}{\mathrm{\π}}_{\mathrm{\α}}^{v_k}=0\\ {\mathrm{\π}}_{\mathrm{\α}}^{v_k}e=1\\ A_{\mathrm{\α}}^{v_k}e=0\end{array}\left(14\right)$

Step 4, utilize formula (15) obtain+1 tactful v of kth_k+1:

$v_{k+1}\∈\arg \underset{v\∈\mathrm{\Ω}}{min}\{A_{\mathrm{\α}}^v{g}_{\mathrm{\α}}^{v_k}+f_{\mathrm{\α}}^v\}---\left(15\right)$

Step 5, utilize formula (11) obtain+1 tactful v of kth_k+1Performance potential vector

Step 6, judgementWhether set up, �� is a less constant; If setting up, then it represents that+1 tactful v of the kth obtained_k+1It is optimal control policy v^*, and exit algorithm, otherwise; After k+1 is assigned to k, return step 4.