CN106027300B - A kind of intelligent robot Parameter Optimization System and method using neural network - Google Patents
A kind of intelligent robot Parameter Optimization System and method using neural network Download PDFInfo
- Publication number
- CN106027300B CN106027300B CN201610343978.7A CN201610343978A CN106027300B CN 106027300 B CN106027300 B CN 106027300B CN 201610343978 A CN201610343978 A CN 201610343978A CN 106027300 B CN106027300 B CN 106027300B
- Authority
- CN
- China
- Prior art keywords
- module
- parameter
- neural network
- compression
- machine people
- 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.)
- Active
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/08—Configuration management of networks or network elements
- H04L41/0803—Configuration setting
- H04L41/0813—Configuration setting characterised by the conditions triggering a change of settings
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/02—Standardisation; Integration
- H04L41/0246—Exchanging or transporting network management information using the Internet; Embedding network management web servers in network elements; Web-services-based protocols
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/08—Configuration management of networks or network elements
- H04L41/0803—Configuration setting
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Manipulator (AREA)
Abstract
The present invention relates to a kind of intelligent robot Parameter Optimization Systems and method using neural network, are related to intelligent robot parameter optimization field.Parameter optimization and compression process are completed by cloud server, to reduce the operand of terminal machine people.The system includes at least one cloud server and at least one terminal machine people, the terminal machine people and cloud server pass through network connection, terminal machine people is internally embedded the terminal neural network module calculated for positive transmission, and cloud server is embedded in the cloud neural network module for optimizing parameter according to sample database.Method includes the following steps: terminal machine people identifies and acquires to environmental information and action command；Different cloud neural network modules carry out parameter optimization to different types of information and store；Judge whether to compress the parameter optimisation procedure of neural network module；Parameter after optimization is compressed.The present invention is suitable for intelligent robot neural network module parameter optimization.
Description
Technical field
The present invention relates to field in intelligent robotics, and in particular to intelligent robot parameter optimization field.
Background technique
Current existing intelligent robot is all single machine version mostly, that is, not by remote server come real
Existing better performance, that there are hardware conditions is bad for this kind of robot, and hardware cost is high, achievable degree of intelligence is not high, uses
Experience bad defect in family.
In order to overcome the shortcomings of standalone version robot, more and more cloud robots gradually appear, and cloud robot is to answer
What the server in cloud was serviced.
Chinese patent is entitled " cloud robot system and implementation method ", publication number: 101973031A, publication date:
2011-02-16 directly passes through network using long-range cloud computing platform and sends control command.
Chinese patent is entitled " a kind of cloud robot system, robot and robot cloud platform ", publication number:
105563484A, publication date: 2016-05-11, with addition of operator attendance modules, and are real by special communication network
Existing data transmission.
Two above patent application has been all made of cloud robot technology, wherein first is carried out eventually by cloud server
End control, second uses cloud server operation mode, can realize network connection, but to the delay requirement of communication network
Very high, when network condition difference, which will lead to the not available situation of service, to be occurred.
Simultaneously as the development of artificial intelligence, the algorithm of machine learning algorithm, especially neural network even depth study is given
Robot industry brings new life, such as convolutional neural networks (CNN), recurrent neural network (RNN) and deep neural network
(DNN) very big progress is brought in speech recognition, visual identity and natural language processing, greatly improves robot product
Application value especially services the user experience of class intelligent robot.
But the advantage of machine learning algorithm is to continue on the parameter that a large amount of data carry out optimization algorithm, for training
Data set substantial amounts, and the quantity of parameter is also especially huge, such as ten several layers of nerve nets applied to visual identity
The parameter of network is 1,000,000,000 orders of magnitude, and the time and calculation resources for completing its training needs are very large.
Such as: a neural network is trained using AlexNet image library, uses dual-socket Xeon
Processors (operational capability of 3TFLOPS) needs 150 hours, even if using the expensive GPU for being exclusively used in machine learning
Server is also required to 2 hours, so realizing that the cost of deep neural network training is very high directly on terminal machine people.
Summary of the invention
Technical problem to be solved by the invention is to provide a kind of intelligent robot parameter optimization systems using neural network
System and method, it is therefore intended that parameter optimization and compression process are completed by cloud server, to reduce the fortune of terminal machine people
Calculation amount.
The technical scheme to solve the above technical problems is that a kind of intelligent robot parameter using neural network
Optimization system, the system comprises at least one cloud server and at least one terminal machine people, the terminal machine people with
Cloud server passes through network connection.
Based on the above technical solution, the present invention can also be improved as follows.
Further, the terminal machine people includes the terminal neural network module calculated for positive transmission.
Further, the terminal machine people further includes the one or more with lower module:
The speech recognition module of voice environment information for identification；
The visual identity module of three-dimensional environment information for identification；
The navigation module of direction environmental information for identification；
For carrying out the route planning module of path planning according to external environment；
For executing the action control module of action command；
The speech recognition module, visual identity module, navigation module, route planning module and action control module difference
Connect terminal neural network module different in cloud server.
Beneficial effect using above-mentioned further scheme is by different terminal neural network modules to terminal machine people
In each module information optimize, reduce operand, improve the accuracy of operation.
Further, the cloud server further include:
For optimizing the cloud neural network module of parameter according to sample database；
For according to the hardware information of the parameter set after optimization and each module in terminal machine people to cloud neural network
The compression module that the parameter optimisation procedure of module is compressed；
Data memory module for being stored to Optimal Parameters.
Further, the compression module further comprises:
For obtain terminal machine people hardware information hardware information obtain module, the hardware information include model,
Computing capability and memory information；
For setting the condition contractive condition setting module for whether carrying out compression of parameters；When compression of parameters condition is between compression
Compression of parameters is carried out when between the upper limit and compression lower limit, the compression upper limit is the maximum operational capability or memory of terminal machine people
Percentage, the compression lower limit is cloud neural network module to the maximum serious forgiveness of compression of parameters；
For parameter digit, the number of plies of neural network, nerve net according to terminal machine people hardware information to neural network
The compression of parameters module that one or more of the node of every layer of network or connection quantity parameter are compressed.
Beneficial effect using above-mentioned further scheme is the bound compressed by setup parameter, is able to ascend user's body
It tests, the operand of cloud server is reduced on the basis of guaranteeing operation accuracy.
Further, the compression module further include:
For being sent to according to compressed parameter and sample database information acquisition estimate error rate, and by the estimate error rate
The error rate budget module of terminal machine people；
For judging whether estimate error rate is greater than the error rate judgment module of the maximum serious forgiveness of compression of parameters, it is, then
It sends a warning message, it is no, then any operation is not made.
Beneficial effect using above-mentioned further scheme is: terminal server can be according to the above process in system software
Corresponding design is made to improve user experience.
Another technical solution that the present invention solves above-mentioned technical problem is as follows: a kind of intelligent robot using neural network
Parameter optimization method, the parameter optimization method the following steps are included:
Step 1: environmental information and action command are identified and are acquired；
It is calculated Step 2: carrying out positive transmission to different types of information；
Step 3: carrying out parameter optimization to different types of information according to sample database and storing；
It is to then follow the steps five Step 4: judge whether to compress the parameter after optimization, it is no, then terminate execution simultaneously
Issue system prompt；
Step 5: being compressed to the parameter after optimization.
Further, the environmental information for identifying and acquiring in step 1 includes voice environment information, three-dimensional environment information and side
To environmental information.
Further, judge whether the process compressed to the parameter after optimization in step 4 are as follows:
Compression of parameters is carried out when compression of parameters condition is between the compression upper limit and compression lower limit, the compression upper limit is
The maximum operational capability of terminal machine people (1) or the percentage of memory, the compression lower limit are that cloud neural network module (3) are right
The maximum serious forgiveness of compression of parameters.
Further, the process parameter after optimization compressed in step 5 are as follows:
Step 5 one, cloud server (2) obtain the hardware information of terminal machine people (1), and the hardware information includes type
Number, computing capability and memory information；
Step 5 two, according to the hardware information of terminal machine people (1) to the parameter digit of neural network, the layer of neural network
One or more of number, the node of every layer of neural network or connection quantity parameter is compressed；
Step 5 three, according to compressed parameter and sample database information acquisition estimate error rate, and by the estimate error rate
It is sent to terminal machine people (1)；
Step 5 four judges whether estimate error rate is greater than the maximum serious forgiveness of compression of parameters, is then to give a warning letter
Breath, it is no, then any operation is not made.
The beneficial effects of the present invention are: the present invention trains neural network by being concentrated use in the high performance service in cloud,
The parameter optimized is sent in all terminal machine people, allows to realize positive fortune by using less hardware resource
It calculates and applies, on the one hand advantageously reduce the hardware cost of terminal machine people and promote user experience, on the other hand pass through concentration
Operation neural network is applied to all terminal machine people, greatly reduce whole system compute repeatedly and operand, and
By the way of non-real-time update parameter, the dependence to networked-induced delay is reduced.
Detailed description of the invention
Fig. 1 is the schematic illustration of the intelligent robot Parameter Optimization System of the present invention using neural network,
The number of middle terminal machine people is one, and includes speech recognition module, visual identity module, navigation in terminal machine people
All modules in module, route planning module and action control module；
Fig. 2 is system principle schematic diagram of the present invention by taking deep neural network as an example, and wherein the number of terminal machine people is
One, and include in the terminal machine people speech recognition module, visual identity module, navigation module, route planning module and
All modules in action control module.
In attached drawing, parts list represented by the reference numerals are as follows:
1, terminal machine people, 2, cloud server, 3, cloud neural network mould it is fast, 4, speech recognition module, 5, vision know
Other module, 6, navigation module, 7, path planning module, 8, action control module, 9, compression module, 10, memory module, 11, end
Terminal nerve network module.
Specific embodiment
The principle and features of the present invention will be described below with reference to the accompanying drawings, and the given examples are served only to explain the present invention, and
It is non-to be used to limit the scope of the invention.
As shown in Figure 1, the invention proposes a kind of intelligent robot Parameter Optimization System and method using neural network,
The parameter optimization is to pass through training data, the process being trained to the parameter in neural network model.Neural network packet
Containing there are many branches, including convolutional neural networks, recurrent neural network and deep neural network, below just with deep neural network
For explanation is further explained to the present invention.
Embodiment 1
As shown in Fig. 2, the system includes at least one cloud server 2 and at least one terminal machine people 1, the terminal
By network connection, terminal machine people 1 is internally embedded the terminal calculated for positive transmission for robot 1 and cloud server 2
Deep neural network module, cloud server 2 are embedded in the cloud depth nerve for optimizing parameter according to sample database
Network module.
Sample database based on when optimizing to parameter can come from terminal machine people, can be from server
Store equipment or network data etc..
The one or more that terminal machine people 1 comprises the following modules:
The speech recognition module 4 of voice environment information for identification；
The visual identity module 5 of three-dimensional environment information for identification；
The navigation module 6 of direction environmental information for identification；
For carrying out the route planning module 7 of path planning according to external environment；
For executing the action control module 8 of action command；
The speech recognition module 4, visual identity module 5, navigation module 6, route planning module 7 and action control module
8 are separately connected terminal deep neural network module different in cloud server 2.
The quasi- parameter optimization that deep neural network is realized with long-range high-performance server of the present invention, the parameter after optimization
Periodically non real-time to update onto terminal machine people, terminal machine people realizes the forward direction of deep neural network with local computing resource
It calculates, relieves the dependence to network.
In order to obtain better parameter optimization effect, the present invention also adds in server 2 beyond the clouds:
For the hardware information according to the parameter set after optimization and each module in terminal machine people 1 to through cloud depth mind
The compression module 9 that parameter after network module optimizes is compressed；
Data memory module 10 for being stored to the parameter after optimization.
Compression module 9 further comprises:
For obtain terminal machine people 1 hardware information hardware information obtain module, the hardware information include model,
Computing capability and memory information；
For setting the condition contractive condition setting module for whether carrying out compression of parameters；When compression of parameters condition is between compression
Compression of parameters is carried out when between the upper limit and compression lower limit, the compression upper limit is the maximum operational capability or interior of terminal machine people 1
The percentage deposited, the compression lower limit are maximum serious forgiveness of the deep neural network module in cloud to compression of parameters；
For according to 1 hardware information of terminal machine people to the parameter digit of deep neural network, the layer of deep neural network
The compression of parameters module that one or more of number, the node of every layer of deep neural network or connection quantity parameter is compressed.
During optimizing parameter, need to simplify the progress of parameter, to every layer of number of nodes according to reality
The parameter situation on border is deleted, as compression of parameters, and cloud server obtains the hardware information of terminal machine people, including type
Number, computing capability and memory information, further according to these information to the parameter digit of deep neural network, deep neural network
The number of plies, the node of every layer of deep neural network and connection quantity optimize, and the parameter after optimization are compressed, compressor bar
The upper limit of part can be the maximum operational capability of terminal machine people or the percentage of memory, and the lower limit of contractive condition is then that parameter is excellent
The maximum serious forgiveness of change, can using the user experience under special scenes it is unaffected be limited set, Cloud Server is excellent
The operational capability and possible mistake that specific data set survey in network after change, and is needed to terminal server report
Accidentally rate needs the information that sounds a warning if error rate exceeds maximum serious forgiveness.
Embodiment 2
A kind of intelligent robot parameter optimization method using neural network is present embodiments provided, equally with depth nerve
Be further described for network, the parameter optimization method the following steps are included:
Step 1: terminal machine people 1 identifies and acquires to environmental information and action command；
It is calculated Step 2: terminal deep neural network module carries out positive transmission to different types of information；
Step 3: cloud deep neural network module carries out parameter optimization to different types of information according to sample database and deposits
Storage；
It is to then follow the steps five Step 4: judge whether to compress the parameter after optimization, it is no, then terminate execution simultaneously
Issue system prompt；
Step 5: compressing to the parameter after optimization, and compressed parameter is sent to terminal machine people 1.
Environmental information includes voice environment information, three-dimensional environment information and direction environmental information.
Judge whether the process compressed to the parameter after optimization in step 4 are as follows:
Compression of parameters is carried out when compression of parameters condition is between the compression upper limit and compression lower limit, the compression upper limit is
The maximum operational capability of terminal machine people 1 or the percentage of memory, the compression lower limit are cloud deep neural network module pair
The maximum serious forgiveness of compression of parameters.
The process that the parameter after optimization is compressed in step 5 are as follows:
Step 5 one, cloud server 2 obtain the hardware information of terminal machine people 1, and the hardware information includes model, meter
Calculation ability and memory information；
Step 5 two, the parameter digit according to the hardware information of terminal machine people 1 to deep neural network, depth nerve net
One or more of the number of plies of network, the node of every layer of deep neural network or connection quantity parameter is compressed；
Step 5 three, according to compressed parameter and sample database information acquisition estimate error rate, and by the estimate error rate
It is sent to terminal machine people 1；
Step 5 four judges whether estimate error rate is greater than the maximum serious forgiveness of compression of parameters, is then to give a warning letter
Breath, it is no, then any operation is not made.
Embodiment 3
The present embodiment for the parameter optimization and compression that apply the present invention to visual identity information to be illustrated.
(1) neural network module for realizing visual identity information is embedded on cloud server；
(2) cloud server is by collecting database, and on network or the terminal machine people in system gets
Tape label image data realizes parameter by specific operation for the input of the node parameter as training neural network
Optimization calculate；
(3) after the calculating of cloud server realization parameter, according to the terminal machine of the result of real network parameter and target
The computing capability of people can choose whether that needing to optimize the positive of whole network calculates, including reduction parameters precision and net
The number of nodes of network and connection, while also can choose and parameter is updated into terminal machine people；
(4) terminal machine people realizes that the positive of neural network calculates using the parameter after optimization, it is also possible to select by
The tape label image locally obtained sends back to cloud server, the image input as the optimization of cloud server next round；
(5) cloud server can also increase certain functions, such as timing, or be opened according to the update status of data set
The parameter optimization of dynamic next round.
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all in spirit of the invention and
Within principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.
Claims (5)
1. a kind of terminal machine people's Parameter Optimization System using neural network, which is characterized in that the system comprises at least one
A cloud server (2) and at least one terminal machine people (1), the terminal machine people (1) and cloud server (2) pass through net
Network connection；
The cloud server (2) further include:
For optimizing the cloud neural network module (3) of parameter according to sample database；
For according to the hardware information of the parameter set after optimization and each module in terminal machine people (1) to through cloud neural network
The compression module (9) that parameter after module (3) optimization is compressed；
Data memory module (10) for being stored to the parameter after optimization；
The compression module (9) further comprises:
Hardware information for obtaining the hardware information of terminal machine people (1) obtains module, and the hardware information includes model, meter
Calculation ability and memory information；
For setting the contractive condition setting module for whether carrying out the condition of compression of parameters；When compression of parameters condition is in compression
Compression of parameters is carried out when between limit and compression lower limit, the compression upper limit is the maximum operational capability or interior of terminal machine people (1)
The available percentage deposited, the compression lower limit are the maximum serious forgiveness of cloud neural network module (3) to compression of parameters；
For parameter digit, the number of plies of neural network, neural network according to terminal machine people (1) hardware information to neural network
The compression of parameters module that one or more of every layer of node or connection quantity parameter are compressed；
The compression module (9) further include:
For being sent to terminal according to compressed parameter and sample database information acquisition estimate error rate, and by the estimate error rate
The error rate budget module of robot (1)；
It is then to issue for judging whether estimate error rate is greater than the error rate judgment module of the maximum serious forgiveness of compression of parameters
Warning message, it is no, then any operation is not made.
2. a kind of terminal machine people's Parameter Optimization System using neural network according to claim 1, which is characterized in that
The terminal machine people (1) includes the terminal neural network module (11) calculated for positive transmission.
3. a kind of terminal machine people's Parameter Optimization System using neural network according to claim 1, which is characterized in that
Terminal machine people (1) further includes with one or more of lower module:
The speech recognition module (4) of voice environment information for identification；
The visual identity module (5) of three-dimensional environment information for identification；
The navigation module (6) of direction environmental information for identification；
For carrying out the route planning module (7) of path planning according to external environment；
For executing the action control module (8) of action command；
The speech recognition module (4), visual identity module (5), navigation module (6), route planning module (7) and action control
Module (8) is separately connected different terminal neural network modules (11).
4. a kind of terminal machine people's parameter optimization method using neural network, which is characterized in that the parameter optimization method packet
Include following steps:
Step 1: environmental information and action command are identified and are acquired；
It is calculated Step 2: carrying out positive transmission to different types of information；
Step 3: carrying out parameter optimization to different types of information according to sample database and storing；
It is to then follow the steps five Step 4: judge whether to compress the parameter after optimization, it is no, then it terminates execution and issues
System prompt；
Step 5: being compressed to the parameter after optimization；
Judge whether the process compressed to the parameter after optimization in step 4 are as follows:
Compression of parameters is carried out when compression of parameters condition is between the compression upper limit and compression lower limit, the compression upper limit is terminal
The maximum operational capability of robot (1) or the available percentage of memory, the compression lower limit are that cloud neural network module (3) are right
The maximum serious forgiveness of compression of parameters；
The process that the parameter after optimization is compressed in step 5 are as follows:
Step 5 one, cloud server (2) obtain the hardware information of terminal machine people (1), and the hardware information includes model, meter
Calculation ability and memory information；
Step 5 two, according to the hardware information of terminal machine people (1) to the parameter digit of neural network, the number of plies of neural network,
One or more of the node of every layer of neural network or connection quantity parameter are compressed；
Step 5 three is sent according to compressed parameter and sample database information acquisition estimate error rate, and by the estimate error rate
To terminal machine people (1)；
Step 5 four judges whether estimate error rate is greater than the maximum serious forgiveness of compression of parameters, is then to send a warning message, no,
Any operation is not made then.
5. a kind of terminal machine people's parameter optimization method using neural network according to claim 4, which is characterized in that
The environmental information for identifying and acquiring in step 1 includes one in voice environment information, three-dimensional environment information and direction environmental information
Kind is several.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610343978.7A CN106027300B (en) | 2016-05-23 | 2016-05-23 | A kind of intelligent robot Parameter Optimization System and method using neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610343978.7A CN106027300B (en) | 2016-05-23 | 2016-05-23 | A kind of intelligent robot Parameter Optimization System and method using neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106027300A CN106027300A (en) | 2016-10-12 |
| CN106027300B true CN106027300B (en) | 2019-03-05 |
Family
ID=57096229
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610343978.7A Active CN106027300B (en) | 2016-05-23 | 2016-05-23 | A kind of intelligent robot Parameter Optimization System and method using neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106027300B (en) |
Families Citing this family (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106873585B (en) * | 2017-01-18 | 2019-12-03 | 上海器魂智能科技有限公司 | A kind of navigation method for searching, robot and system |
| CN106972959A (en) * | 2017-03-15 | 2017-07-21 | 千寻位置网络有限公司 | A kind of online service method and system of sensing algorithm |
| CN109102074B (en) * | 2017-06-21 | 2021-06-01 | 上海寒武纪信息科技有限公司 | Training device |
| US11321604B2 (en) * | 2017-06-21 | 2022-05-03 | Arm Ltd. | Systems and devices for compressing neural network parameters |
| CN109685088A (en) * | 2017-10-18 | 2019-04-26 | 上海仪电（集团）有限公司中央研究院 | Narrow band communication intelligent image analysis system based on cloud separation convolutional neural networks |
| CN107766889B (en) * | 2017-10-26 | 2021-06-04 | 浪潮集团有限公司 | Cloud edge computing fused deep learning computing system and method |
| CN107797867A (en) * | 2017-10-31 | 2018-03-13 | 济南浪潮高新科技投资发展有限公司 | A kind of method and device for strengthening edge side intelligence computation ability |
| WO2019183568A1 (en) | 2018-03-23 | 2019-09-26 | Google Llc | Controlling a robot based on free-form natural language input |
| CN109543829A (en) * | 2018-10-15 | 2019-03-29 | 华东计算技术研究所（中国电子科技集团公司第三十二研究所） | Method and system for hybrid deployment of deep learning neural network on terminal and cloud |
| US11584020B2 (en) * | 2018-12-04 | 2023-02-21 | Cloudminds Robotics Co., Ltd. | Human augmented cloud-based robotics intelligence framework and associated methods |
| CN109902818B (en) * | 2019-01-15 | 2021-05-25 | 中国科学院信息工程研究所 | Distributed acceleration method and system for deep learning training task |
| CN109746918B (en) * | 2019-02-14 | 2021-08-13 | 中山大学 | Optimization method for delay of cloud robot system based on joint optimization |
| CN109993308B (en) * | 2019-03-29 | 2024-05-24 | 深圳先进技术研究院 | Cloud platform-based shared learning system and method, shared platform and method and medium |
| CN111954206B (en) * | 2019-05-17 | 2024-04-09 | 株式会社Ntt都科摩 | Terminal and base station |
| CN110427536B (en) * | 2019-08-12 | 2022-03-04 | 深圳忆海原识科技有限公司 | Brain-like decision and motion control system |
| CN113076965A (en) * | 2020-01-06 | 2021-07-06 | 广州中国科学院先进技术研究所 | Cloud-based service robot scene classification system and method |
| CN113470653A (en) * | 2020-03-31 | 2021-10-01 | 华为技术有限公司 | Voiceprint recognition method, electronic equipment and system |
| CN113222530A (en) * | 2021-04-22 | 2021-08-06 | 北京洛必德科技有限公司 | Robot and environment cooperative work system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101973031A (en) * | 2010-08-24 | 2011-02-16 | 中国科学院深圳先进技术研究院 | Cloud robot system and implementation method |
| CN102253644A (en) * | 2010-05-17 | 2011-11-23 | 无锡爱德普信息技术有限公司 | Configurable intelligent process control system and implementation method thereof |
| CN105550713A (en) * | 2015-12-21 | 2016-05-04 | 中国石油大学(华东) | Video event detection method of continuous learning |
| CN105563484A (en) * | 2015-12-08 | 2016-05-11 | 深圳前海达闼云端智能科技有限公司 | Cloud robot system, robot and robot cloud platform |
-
2016
- 2016-05-23 CN CN201610343978.7A patent/CN106027300B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102253644A (en) * | 2010-05-17 | 2011-11-23 | 无锡爱德普信息技术有限公司 | Configurable intelligent process control system and implementation method thereof |
| CN101973031A (en) * | 2010-08-24 | 2011-02-16 | 中国科学院深圳先进技术研究院 | Cloud robot system and implementation method |
| CN105563484A (en) * | 2015-12-08 | 2016-05-11 | 深圳前海达闼云端智能科技有限公司 | Cloud robot system, robot and robot cloud platform |
| CN105550713A (en) * | 2015-12-21 | 2016-05-04 | 中国石油大学(华东) | Video event detection method of continuous learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106027300A (en) | 2016-10-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106027300B (en) | A kind of intelligent robot Parameter Optimization System and method using neural network | |
| CN107181818B (en) | Robot remote control and management system and method based on cloud platform | |
| CN107194612B (en) | Train operation scheduling method and system based on deep reinforcement learning | |
| CN109819233A (en) | A kind of digital twinned system based on virtual image technology | |
| CN109271374A (en) | A kind of database health scoring method and scoring system based on machine learning | |
| CN114442510A (en) | Digital twin closed loop control method, system, computer device and storage medium | |
| CN109993734A (en) | Method and apparatus for output information | |
| CN108764456B (en) | Airborne target identification model construction platform, airborne target identification method and equipment | |
| CN109990790A (en) | A kind of unmanned plane paths planning method and device | |
| CN106373160B (en) | A kind of video camera active target localization method based on deeply study | |
| CN109657077A (en) | Model training method, lane line generation method, equipment and storage medium | |
| CN112541584B (en) | Deep neural network model parallel mode selection method | |
| CN105340311B (en) | The network equipment, network service prediction meanss and method | |
| CN113037783B (en) | Abnormal behavior detection method and system | |
| CN117035376B (en) | Dynamic monitoring method, system and medium for emergency rescue exercise | |
| CN110992307A (en) | Insulator positioning and identifying method and device based on YOLO | |
| CN109919302A (en) | A kind of neural network training method and device | |
| CN109685805A (en) | A kind of image partition method and device | |
| CN111914600A (en) | Group emotion recognition method based on space attention model | |
| CN109931940A (en) | A kind of robot localization method for evaluating confidence based on monocular vision | |
| CN112884164A (en) | Federal machine learning migration method and system for intelligent mobile terminal | |
| CN108124479A (en) | Map labeling method and device, cloud server, terminal and application program | |
| CN112395952A (en) | A unmanned aerial vehicle for rail defect detection | |
| CN115952989A (en) | Natural disaster emergency scheduling system based on digital twins | |
| CN112486667B (en) | Method and device for accurately processing data based on edge calculation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |