CN108582072A - A kind of space manipulator mission planning method based on improvement figure planning algorithm - Google Patents
A kind of space manipulator mission planning method based on improvement figure planning algorithm Download PDFInfo
- Publication number
- CN108582072A CN108582072A CN201810403035.8A CN201810403035A CN108582072A CN 108582072 A CN108582072 A CN 108582072A CN 201810403035 A CN201810403035 A CN 201810403035A CN 108582072 A CN108582072 A CN 108582072A
- Authority
- CN
- China
- Prior art keywords
- planning
- matrix
- state
- task
- mechanical arm
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1656—Programme controls characterised by programming, planning systems for manipulators
- B25J9/1661—Programme controls characterised by programming, planning systems for manipulators characterised by task planning, object-oriented languages
Landscapes
- Engineering & Computer Science (AREA)
- Robotics (AREA)
- Mechanical Engineering (AREA)
- Manipulator (AREA)
- Numerical Control (AREA)
Abstract
An embodiment of the present invention provides a kind of space manipulator mission planning method based on modified figure planning algorithm, realize that mechanical arm is directed to the mission planning of different task scene, including：For operation task, environment and robotic arm manipulation ability, the general mathematical model of mechanical arm mission planning is established；On the basis of task mathematical model, improves figure using the simulated annealing for introducing comentropy and expand, Expansion Planning disaggregation while control figure expands rate of convergence；According to specific action planning algorithm, fusion different target is completed in constraint space, is completed the extraction of mechanical arm mission planning solution, is realized mechanical arm mission planning.The technical solution provided according to embodiments of the present invention is, it can be achieved that consider the autonomous mission planning of space manipulator of multiple-objection optimization under multi-constraint condition.
Description
【Technical field】
The present invention relates to a kind of space manipulator mission planning methods that multiple-objection optimization is considered under multi-constraint condition, belong to
Mechanical arm mission planning field.
【Background technology】
As the current international rapid development of aerospace industry and deepening continuously for space probation, spatial operation task also become
It becomes increasingly complex.Substitute or assist astronaut to complete the in-orbit operation task of all kinds of complexity using space manipulator, it can be effective
Economic benefit and the safety of spacecraft in-orbit service are promoted, therefore, space manipulator will complete autonomous mission planning in future
Occupy very important status in space exploration.
Acquisition of information difficulty, uncertain factor are more in space working environment, the remote operating time delay of spacecraft and command centre
Problem is more prominent, this proposes the ability of making decisions on one's own of space manipulator new requirement.Meanwhile being limited to complicated space
The operation task of environment and operation task, space manipulator has very strong binding character, such as this kind of unplanned for space maintenance
It is interior and cannot by single action realize task be needed first under the premise of meeting such as joint angles, angular speed constraint
The new component of installation could be replaced by removing worn part.Further, since the finiteness of space resources, for the rational of spacecraft resource
Configuration is also particularly important.Therefore, research considers the autonomous mission planning method of multiple-objection optimization to future under multi-constraint condition
Chinese Space Manipulator Technology development has important practical significance in space exploration.
【Invention content】
In view of this, an embodiment of the present invention provides a kind of, the space manipulator task based on modified figure planning algorithm is advised
The method of drawing, to realize the autonomous mission planning of mechanical arm for considering multiple-objection optimization under multi-constraint condition.
The above-mentioned space manipulator mission planning method based on modified figure planning algorithm includes at least：
According to specific tasks scene, mathematical description is carried out to task status and task action respectively, establishes mechanical arm task
The mathematical model of planning；
According to the simulated annealing for introducing comentropy, realize that planning chart continues to expand after dbjective state appearance, expands
Go out the planning solution other than most short action sequence, so that figure planning is jumped out local optimum, realize the acquisition of globally optimal solution, complete to figure
The improvement of expansion；
It is expanded according to figure as a result, planning the solution stage using search strategy from the front to the back in extraction, and fully consider task
Constraint merges different task optimization targets under the premise of meeting constraint, realizes and merges multitask target in constraint space
Mission planning solution extraction.
In the above method, the mechanical arm mission planning mathematical model includes：
The first step, 4 × 4 homogeneous transform matrix T according to D-H parametric methods structure description mechanical arm tail end poses.For
Task environment, constructing environment state matrix WS, form is as follows：
Wherein,To operate the location expression matrix of object,State for operation object and mechanical arm meets matrix, and i-th
Row j column elements indicate matching degree of the object to certain possible state：1 representative meets the state, and 0 representative is not met, to ensure square
Battle array it is homogeneity, with zero row or zero row polishing matrix.Concrete form difference is as follows：
Assuming that there are p operable objects in environment, operating object has q kind possible states, then n=max { p, q }.
Comprehensive mechanical arm state and task environment state, task status can be described as matrix S, and matrix form is as follows：
Second step, definition action matrix A describe task action, and matrix A form is as follows：
Wherein, TAFor description the translating of mechanical arm tail end coordinate system, 4 × 4 rotationally-varying homogeneous matrix, WAFor description
The matrix of object state change, W are operated in environmentAFollowing form can be decomposed into：
Wherein,To describe 4 × 4 homogeneous matrix of object translation in environment, rotation,Indicate i-th of object shape
The homogeneous matrix of n × n of state variation, can be obtained by the elementary rank transform of diagonal matrix I.
Third walks, and by the interaction of matrix, the detailed description of mission planning process may be implemented.If mission planning
The state matrix of -1 step of kth is Sk-1, action matrix is Ak-1, multiplication can obtain：
Wherein, STransThe state-transition matrix of mechanical arm, W after being acted for executionTransState to describe environment shifts square
Battle array, is multiplied to obtain, can be indicated by following matrixs in block form by matrix in block form：
Wherein,It is arranged for 4 × n of n rows by action matrix effect, the matrix of n rows n × n row, respectively
The matrix that the location matrix of operation object is multiplied with the state matrix of characterization behaviour's crop condition in characterization environment.It can therefrom carry
Take out the task status S of current time stepk, specific element relation is as follows：
Process is established by above-mentioned mathematical model, realizes and is described for the task process of different mechanical arms and environment,
Solution for the task execution process of multi-restriction Multi-Objective is laid a good foundation.
In the above method, the improvement that described pair of figure is expanded includes：
The first step：The comentropy for the state that current time walks in figure expanding course is defined as：
Wherein pkIt indicates kth class state proportion in time step t, is represented by：Q is in current time step
The quantity summation of all states, k=0 represent the state occurred in any time before walks, number of states q0, k=1
Represent the state not occurred in any time before walks, number of states q1。
Using the comentropy of each time step as the attenuation coefficient of temperature during simulated annealing, and using comentropy as ratio
Decaying, is shown below：
Wherein rtIndicate temperature decline coefficient, TtIndicate temperature of the simulated annealing in time step t,Expression is based on
The temperature for the time step t+1 that comentropy is found out.
Under normal conditions, it is contemplated that the validity of simulated annealing, and to ensure the stability of simulated annealing,
The temperature of its single step unconfined cannot decay.For the situation, the single step minimum attenuation temperature of simulated annealing is defined, is advised
Its fixed temperature drop rate should be no faster than Tt min, Tt minForm is as follows：
Tt min=T0/(1+ln(t-t0+1)) (12)
Wherein Tt minIndicate algorithm in the minimum temperature of time step t, T0For the initial temperature of simulated annealing, t0It indicates
Search reaches step number used in original state for the first time.
Then by temperature damping's process, the self-adjusting simulated annealing according to figure planning information may be implemented, change
Temperature damping's process into simulated annealing can be expressed as：
Second step：Scheme to expand efficiency to improve, planning chart is inversely built from task object state, according to the mesh of task
State set is marked, is expanded by the behavior aggregate of task.When figure planning is not up to original state, figure, which is expanded, to be counted as making interior energy
Increase process, figure is expanded to be continued according to normal search process.After searching out original state, when planning chart reaches just
After beginning state, figure expands the process for being considered as making interior energy to reduce, and figure is expanded to be continued to expand with certain probability, and this
Probability is continuously decreased with the passage of time step, stops expanding until reaching setting value.This probability is represented by：
Wherein, dE is the energy difference of transfer, and T is Current Temperatures, and k is proportionality coefficient.Due to being annealing process, so dE<
0, temperature T is higher, and state transition probability P (dE) is bigger.
Improved figure is expanded schematic diagram and is please referred to Fig.2.
In the above method, the mission planning solution extraction that multiple target is merged in constraint space includes
During specific action planning algorithms selection, it is based on different task demand, needs to consider corresponding task index, such as
Joint stroke, end distance, flexibility ratio etc..By setting corresponding weighted value, different task target is merged, mission planning is extracted
Solution.Kth step action cost can be calculated by following formula in extraction process：
fk(p)=ω1f1(p)+ω2f2(p)…+ωifi(p)+… (15)
Wherein, fiFor the calculating function for different task target designed according to physical planning algorithm, ωiFor corresponding mesh
Target cost weighted value, p indicate the decision variable in planning process, such as joint angle, terminal position posture, can be by one group of collection
It closes and indicates：
P={ p1,p2,p3,...} (16)
By the above process, corresponding task object is calculated for different decision variables to may be implemented by Weighted Fusion
For the action cost calculating process of multiple target.
The constraint space of space manipulator operation task is built based on corresponding task restriction, and is limited according to specific
Value builds corresponding constraint subspace.In acting selection course, need to ensure that the value of each decision variable in planning process is in
Present confinement subspace CnowIn, it is shown below：
p1,p2,p3...∈Cnow (17)
Under the premise of meeting task restriction, selection minimizes object function fk(p) action planning algorithm, realization action
The preferred process of planning algorithm.
Specific extraction step is as follows：
The first step finds the proposition layer for including all original states in planning chart, and builds state using original state
Set；
Second step finds the set of actions that set is realized by additive effect in last layer acts layer；
Third walks, and action physical planning algorithm is selected according to action node；
4th step finds a layer state set thereon according to the precondition of set of actions；
5th step turns the 6th step and otherwise goes to second step if including dbjective state in state set；
6th step turns the 7th step and otherwise turns the first step if all proposition layers comprising original state of traversal；
7th step solves cost value optimum programming as a result, planning terminates.
The mechanical arm towards different task can be realized by the process according to different task Objective extraction mission planning solution
Mission planning process.
The technical solution of the embodiment of the present invention has the advantages that：
(1) present invention is directed to the task characteristic of space manipulator, realizes the mathematical description of mechanical arm mission planning, solves
The problem of mechanical arm mission planning mathematical description difficulty realizes mission planning process and is turned by abstractdesription to mathematical description
Change；
(2) present invention introduces comentropy in simulated annealing, and dbjective state during figure is expanded is stated by comentropy
Information content so that effective combination that enhanced simulated annealing is expanded with figure realizes the expansion of task action sequence sets；
(3) present invention, according to specific action layer planning algorithm, examines on the basis of improving figure expansion in constraint space
Consider fusion multitask target, realizes the preferred of planning solution.
Advantage according to the present invention can be realized and cutd open to task on the basis of mechanical arm mission planning mathematical description
Surface analysis obtains mission planning disaggregation, considers task restriction and task object during task execution, realizes that mechanical arm can be held
Row strategy it is preferred.The technology can be applied to consider under multi-constraint condition the space manipulator mission planning of multiple-objection optimization
Journey.
【Description of the drawings】
In order to illustrate the technical solution of the embodiments of the present invention more clearly, below will be to needed in the embodiment attached
Figure is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention, for this field
For those of ordinary skill, without having to pay creative labor, it can also be obtained according to these attached drawings other attached
Figure.
Fig. 1 is the mechanical arm mission planning flow based on modified figure planning algorithm that embodiment of the present invention is provided
Figure；
Fig. 2 is that the figure designed in the embodiment of the present invention expands flow chart；
It is figure that Fig. 3, which is that research object is built in the embodiment of the present invention,.
【Specific implementation mode】
1, the kinematics model of Five-degree-of-freedom spatial mechanical arm is established as shown in figure 3, its corresponding D-H parameter such as 1 institute of table
Show.The operation console of the accurate operations such as the mechanical arm and executable jack, displacement forms mechanical arm system, and industrial personal computer passes through CAN cards
Corresponding joint of mechanical arm is connected, execution driving corresponding joint movement is issued, realizes the implementation procedure for corresponding task.
The corresponding D-H parameters of 1 Five-degree-of-freedom spatial mechanical arm initial configuration of table
d1=97.1, d2=273.4, d3=560, d4=550, d5=126.6, d6=224.7
2, mechanical arm is chosen based on above-mentioned mechanical arm task mathematics description method for the displacement task of two objects
Crawl, mobile and release movement build task action collection, and the precondition each acted, additive effect is arranged and deletes effect.
Such as：The precondition of grasping movement is that mechanical arm tail end position is overlapped with object space and robot arm end effector opens, and is added
Add effect be mechanical arm capture position and object be crawled position 1 and robot arm end effector is closed, it is machinery to delete effect
It is 0 and robot arm end effector opening that arm, which captures position and the position that is crawled of object,.
3, according to the working range of mechanical arm, rational operation object location is selected, setting mechanical arm initial configuration is initial
It is configured as C=[- 1.788 °, -69.394 °, 115.074 °, 44.320 °, 0 °] T, the initial position A1=of operation object A [-
0.08m, 0.8m, 0.05m] T, it is expected that position A2=[0.3m, 0.7m, the 0.05m] T being transferred to, the initial position of operation object B
B1=[- 0.02m, 0.7m, 0.05m] T, it is expected that position B2=[0.5m, 0.6m, the 0.05m] T being transferred to.It is planned with straight line
Based on, plan corresponding task, in planning process, crawl and release movement respectively include it is downward plan with upward straight line with
And the closure and separate operation of paw, mechanical arm movement have the abilities such as straight line planning and the collision-free Trajectory Planning of Welding based on A*.
4, it is expanded, task object state is expanded, directly based on the improved figure of simulated annealing for introducing comentropy
Reach setting value to Simulated annealing, obtains and advised by terminating to original state since dbjective state for the task of action connection
Draw figure.
5, joint angle and joint angular speed constraint are set in task planning process, and with the pass during manipulator motion
Section stroke is that optimization aim inversely extracts action sequence by the physical planning algorithm of action in planning chart with end distance
Planning solution selects OPTIMAL TASK implementation strategy according to corresponding optimization aim.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention
With within principle, any modification, equivalent substitution, improvement and etc. done should be included within the scope of protection of the invention god.
The content that description in the present invention is not described in detail belongs to the known technology of those skilled in the art.
Claims (4)
1. a kind of based on the space manipulator mission planning method for improving figure planning algorithm, it is characterised in that this method includes following
Step：
(1) according to specific tasks scene, mathematical description is carried out to task status and task action, establishes mechanical arm mission planning
Mathematical model；
(2) simulated annealing according to the mathematical model of the mission planning and introducing comentropy, design drawing are expanded, and are obtained and are appointed
Business planning chart；
(3) according to the mission planning figure, different task target is merged in constraint space, in conjunction with specific action planning algorithm,
It realizes the extraction of mission planning solution, completes the mission planning of space manipulator.
2. according to the method described in claim 1, which is characterized in that the mathematical model of the mechanical arm mission planning is at least wrapped
It includes：
(1) according to specific tasks scene, task status matrix S is defined, form is as follows：
Wherein TsFor the description using D-H parametric methods to mechanical arm tail end pose, WSIt is as follows for ambient condition matrix form：
Wherein,To operate the location expression matrix of object,Meet matrix, the i-th row j row for the state of operation object and mechanical arm
Matching degree of the element representation object to certain possible state：1 representative meets the state, and 0 representative is not met, to ensure the neat of matrix
Secondary property, with zero row or zero row polishing matrix.Concrete form difference is as follows：
Assuming that there are p operable objects in environment, operating object has q kind possible states, then n=max { p, q }；
(2) according to operational motion, task action matrix A is defined, form is as follows：
Wherein, TAFor description the translating of mechanical arm tail end coordinate system, 4 × 4 rotationally-varying homogeneous matrix, WATo describe environment
The matrix of middle operation object state change, WAFollowing form can be decomposed into：
Wherein,To describe 4 × 4 homogeneous matrix of object translation in environment, rotation,Indicate that i-th of object state becomes
The homogeneous matrix of n × n of change can be obtained by the elementary rank transform of diagonal matrix I；
(3) according to state matrix S defined above and action matrix A, definition status transfer matrix STrans, form is as follows：
Wherein, STransThe state-transition matrix of mechanical arm, W after being acted for executionTransTo describe the state-transition matrix of environment, lead to
It crosses matrix in block form to be multiplied to obtain, can be indicated by following matrixs in block form：
The location matrix of operation object in environment is respectively characterized to be multiplied with the state matrix of characterization behaviour's crop condition
Obtained matrix therefrom can extract out the task status S of current time stepk, specific element relation is as follows：
3. according to the method described in claim 1, which is characterized in that the figure expansion includes at least：
(1) comentropy of current time step state is defined as in figure expanding course：
Wherein pkIt indicates kth class state proportion in time step t, is represented by：Q is to own in current time step
State quantity summation, k=0 represents the state that occurred in any time before walks, number of states q0, k=1 representatives
The state not occurred in any time before walks, number of states q1, moved back the comentropy of each time step as simulation
Temperature T during firetAttenuation coefficient rt, and decay by ratio of comentropy, it is shown below：
WhereinBe attenuation coefficient be rtAt present the temperature of a time step, it is generally the case that provide that its temperature drop rate should not
It is faster thanForm is as follows：
WhereinIndicate algorithm in the minimum temperature of time step t, T0For the initial temperature of simulated annealing, t0It indicates for the first time
Search reaches step number used in original state, then by temperature damping's process, may be implemented adaptive according to figure planning information
Temperature damping's process of simulated annealing, modified-immune algorithm can be expressed as：
(2) it is to improve figure to expand efficiency, from the reverse structure planning chart of task object state, according to the dbjective state of task
Collection, is expanded by the behavior aggregate of task, and when figure planning is not up to original state, figure, which is expanded, is counted as that interior energy is made to increase
Journey, figure is expanded to be continued according to traditional search process, after searching out original state, by the state in figure expanding course
Temperature decline coefficient of the comentropy as simulated annealing, and then realize that the self-adaptive temperature of simulated annealing process was decayed
Journey, figure, which is expanded, after reaching expansion end condition terminates, and obtains mission planning figure.
4. according to the method described in claim 1, which is characterized in that the mission planning solution extraction includes at least：
(1) according to different task demand, planning solves kth step action cost in extraction process and is represented by：
fk(p)=ω1f1(p)+ω2f2(p)…+ωifi(p)+…
Wherein fiFor the calculating function for different task target designed according to physical planning algorithm, p is that the decision in task becomes
Amount, such as joint angle, terminal position posture, ωiFor weighted value, different task target is merged to realize；
(2) according to specific task restriction limit value, corresponding constraint subspace is built, specific action planning algorithm is selected,
Such as A* obstacle-avoiding route plannings algorithm, Descartes's straight line are planned, ensure that each decision variable is in constraint subspace in planning process
In；
(3) according to mission planning figure, reverse extraction planning solution traverses all state layers for including original state, is meeting task
Under the premise of constraint, selection minimizes object function fk(p) multiple target is merged in action planning algorithm, realization in constraint space
Mission planning solution extraction, complete mechanical arm mission planning.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810403035.8A CN108582072B (en) | 2018-04-28 | 2018-04-28 | Improved graph planning algorithm-based space manipulator task planning method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810403035.8A CN108582072B (en) | 2018-04-28 | 2018-04-28 | Improved graph planning algorithm-based space manipulator task planning method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108582072A true CN108582072A (en) | 2018-09-28 |
| CN108582072B CN108582072B (en) | 2020-09-15 |
Family
ID=63620078
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810403035.8A Active CN108582072B (en) | 2018-04-28 | 2018-04-28 | Improved graph planning algorithm-based space manipulator task planning method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108582072B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110514567A (en) * | 2019-08-28 | 2019-11-29 | 哈尔滨工业大学 | Gas source searching method based on comentropy |
| CN115610704A (en) * | 2022-09-27 | 2023-01-17 | 哈尔滨工业大学 | Orbital transfer method, device and medium capable of realizing grazing flight observation task on orbit |
| CN116673968A (en) * | 2023-08-03 | 2023-09-01 | 南京云创大数据科技股份有限公司 | Mechanical arm track planning element selection method and system based on reinforcement learning |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103576680A (en) * | 2012-07-25 | 2014-02-12 | 中国原子能科学研究院 | Robot path planning method and device |
| CN106338966A (en) * | 2016-10-27 | 2017-01-18 | 福州大学 | Novel programming method for trajectory planning of industrial robot |
| CN106444738A (en) * | 2016-05-24 | 2017-02-22 | 武汉科技大学 | Mobile robot path planning method based on dynamic motion primitive learning model |
| CN106527424A (en) * | 2016-09-20 | 2017-03-22 | 深圳市银星智能科技股份有限公司 | Mobile robot and navigation method of mobile robot |
| CN107544502A (en) * | 2017-09-25 | 2018-01-05 | 华中科技大学 | A kind of Planning of Mobile Robot under known environment |
-
2018
- 2018-04-28 CN CN201810403035.8A patent/CN108582072B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103576680A (en) * | 2012-07-25 | 2014-02-12 | 中国原子能科学研究院 | Robot path planning method and device |
| CN106444738A (en) * | 2016-05-24 | 2017-02-22 | 武汉科技大学 | Mobile robot path planning method based on dynamic motion primitive learning model |
| CN106527424A (en) * | 2016-09-20 | 2017-03-22 | 深圳市银星智能科技股份有限公司 | Mobile robot and navigation method of mobile robot |
| CN106338966A (en) * | 2016-10-27 | 2017-01-18 | 福州大学 | Novel programming method for trajectory planning of industrial robot |
| CN107544502A (en) * | 2017-09-25 | 2018-01-05 | 华中科技大学 | A kind of Planning of Mobile Robot under known environment |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110514567A (en) * | 2019-08-28 | 2019-11-29 | 哈尔滨工业大学 | Gas source searching method based on comentropy |
| CN110514567B (en) * | 2019-08-28 | 2021-10-29 | 哈尔滨工业大学 | Gas source searching method based on information entropy |
| CN115610704A (en) * | 2022-09-27 | 2023-01-17 | 哈尔滨工业大学 | Orbital transfer method, device and medium capable of realizing grazing flight observation task on orbit |
| CN115610704B (en) * | 2022-09-27 | 2023-09-29 | 哈尔滨工业大学 | Rail changing method, device and medium capable of realizing glancing observation task on rail |
| CN116673968A (en) * | 2023-08-03 | 2023-09-01 | 南京云创大数据科技股份有限公司 | Mechanical arm track planning element selection method and system based on reinforcement learning |
| CN116673968B (en) * | 2023-08-03 | 2023-10-10 | 南京云创大数据科技股份有限公司 | Mechanical arm track planning element selection method and system based on reinforcement learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108582072B (en) | 2020-09-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Arndt et al. | Meta reinforcement learning for sim-to-real domain adaptation | |
| Sadeghi et al. | Sim2real viewpoint invariant visual servoing by recurrent control | |
| CN108582072A (en) | A kind of space manipulator mission planning method based on improvement figure planning algorithm | |
| CN109960880B (en) | Industrial robot obstacle avoidance path planning method based on machine learning | |
| CN111273664B (en) | Multi-robot path collaborative planning method for space truss structure on-orbit assembly | |
| CN111240356B (en) | Unmanned aerial vehicle cluster convergence method based on deep reinforcement learning | |
| Harada et al. | A manipulation motion planner for dual-arm industrial manipulators | |
| CN104331547A (en) | Space mechanical arm structure parameter optimization method based on operability | |
| Yue et al. | Experimental research on deep reinforcement learning in autonomous navigation of mobile robot | |
| CN109901603B (en) | Multi-spacecraft attitude cooperative control method under input time delay | |
| Hussein et al. | Deep reward shaping from demonstrations | |
| Jin et al. | Distributed competition of multi-robot coordination under variable and switching topologies | |
| Liu et al. | A reinforcement learning-based framework for robot manipulation skill acquisition | |
| CN107957895A (en) | Agile coordination control strategy for earth satellite | |
| Raessa et al. | Teaching a robot to use electric tools with regrasp planning | |
| Lima et al. | A probabilistic cellular automata ant memory model for a swarm of foraging robots | |
| Oyama et al. | Inverse kinematics learning for robotic arms with fewer degrees of freedom by modular neural network systems | |
| Desai et al. | Interactive co-design of form and function for legged robots using the adjoint method | |
| Takayanagi et al. | Hierarchical task planning from object goal state for human-assist robot | |
| CN113534660A (en) | Multi-agent system cooperative control method and system based on reinforcement learning algorithm | |
| Hiranaka et al. | Primitive Skill-based Robot Learning from Human Evaluative Feedback | |
| Lima et al. | Formal analysis in a cellular automata ant model using swarm intelligence in robotics foraging task | |
| Tang et al. | Coordinated motion planning of dual-arm space robot with deep reinforcement learning | |
| CN115327926A (en) | Multi-agent dynamic coverage control method and system based on deep reinforcement learning | |
| Hong et al. | Deterministic policy gradient based formation control for multi-agent systems |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |