Manipulator execution path planning method for complex space guided measurement task

A manipulator execution path planning method for a complex space guided measurement task, comprising the following steps:

• S1, determining a measurement space Ω of a manipulator;
• S2, determining a starting point and a target point of a manipulator path;
• S3, using a collision algorithm, determining an obstructed space Ω0 and a free space Ωf of the measurement space Ω;
• and determining a proportion of the obstructed space in the measurement space;
• wherein a calculation formula for the proportion is expressed as:
• λ
• =
• Ω
• 0
• /
• Ω
• ,
• λϵ
• (
• 0
• ,
• 1
• ]
• ;
• S4, constructing a space compression model to compress the measurement space and remove an invalid space in the measurement space;
• wherein S4 specifically comprises:
• when λ>0.5, constructing a first model to compress the measurement space;
• and when λ≤0.5, constructing a hyper-ellipsoidal model to compress the measurement space;
• wherein the first model is expressed as:
• {
• x
• =
• k
• 1
• ·
• t
• y
• =
• k
• 2
• ·
• [
• (
• x
• -
• k
• 3
• )
• +
• k
• 4
• ]
• +
• t
• ·
• ε
• or
• {
• x
• =
• k
• 1
• ·
• t
• y
• =
• k
• 2
• ·
• [
• (
• x
• -
• k
• 3
• )
• +
• k
• 4
• ]
• +
• 2
• ·
• t
• ·
• ε
• z
• =
• k
• 5
• ·
• [
• (
• x
• -
• k
• 3
• )
• +
• k
• 6
• ]
• +
• 2
• ·
• t
• ·
• ε
• ;
• wherein k1 is a boundary value of the measurement space, and k1=maxΩ−minΩ;
• k2, k3, k4, k5 and k6 are constants related to the starting point and the target point;
• t∈(0,1) is a random number conforming to uniform distribution;
• ε is a compression constant, and ε=γλ, wherein γ is a constant;
• the hyper-ellipsoidal model is expressed as:
• F
• ⁡
• (
• x
• ,
• y
• ,
• z
• )
• =
• (
• x
• -
• x
• 0
• )
• 2
• a
• 0
• 2
• +
• (
• y
• -
• y
• 0
• )
• 2
• a
• 1
• 2
• +
• (
• z
• -
• z
• 0
• )
• 2
• a
• 2
• 2
• -
• 1
• ;
• wherein a0 is a long axis, a0=∥xi−xg∥;
• xi is the starting point;
• xg is the target point;
• a1 is a medial axis,
• a
• 1
• =
• c
• ⁡
• (
• t
• )
• 2
• -
• a
• 0
• 2
• 2
• 2
• ;
• a2 is a short axis,
• a
• 2
• =
• c
• ⁡
• (
• t
• )
• 2
• -
• a
• 0
• 2
• 2
• 4
• ;
• c(t) is a path length obtained last time;
• and (x0, y0, z0) is coordinates of the starting point;
• S5, in the compressed measurement space Ωn, finding candidate paths for connecting the starting point and the target point by constructing an adjacent node tree connected topology network according to a preset iterative step length, and constructing a candidate path dataset by using all the candidate paths;
• S6, establishing a reinforcement learning drive model for the candidate path dataset, and finding a required path by utilizing a target equation with the shortest path and time;
• S7, performing high-order curve fitting on the required path to achieve the smoothing of the required path and obtain a final path;
• S8, discretizing the final path obtained in step S7, solving and calculating joint angles of the manipulator by using an inverse kinematics algorithm, and determining an angle of each joint of the manipulator to obtain a joint path of each joint moving in a space;
• and
• S9, calculating whether interference exists between each joint path and the obstructed space;
• when the interference exists between a joint path and the obstructed space, adjusting parameters of the space compression model, and returning to S4 to be re-performed;
• and when no interference exists between any joint path and the obstructed space, completing path control from the starting point to the target point on the manipulator according to the joint path of each joint moving in the space in S8.