Robustness simulation of control algorithm for human-simulated intelligence based fusion*

Shun-lan ZHU, Hui HUANG, Si-ping ZHANG

1School of Automation, Chongqing Industry Polytechnic College, Chongqing 401120, China;2CISDI (Group) Electric Technology Co., Ltd., Chongqing 400013，China



Robustness simulation of control algorithm for human-simulated intelligence based fusion*

Shun-lan ZHU† 1, Hui HUANG1, Si-ping ZHANG2

1School of Automation, Chongqing Industry Polytechnic College, Chongqing 401120, China;2CISDI (Group) Electric Technology Co., Ltd., Chongqing 400013，China

Abstract:In order to overcome that it is difficult to realize the control resulted in poor robustness of control algorithm for complex correlation system with uncertainty, the paper explored the robustness of human-simulated intelligence based fusion control algorithm. In the paper, it discussed the facing challenge of conventional control, analyzed the cybernetics characteristic of complex correlation system with uncertainty, researched on control strategy of human-simulated intelligence based fusion control strategy, and aimed at the specific object it constructed the corresponding control algorithm. It took a two order object with time lag as an example, carried through the simulation contrast research respectively by conventional PID and human simulated based fusion control algorithm, and the experiment response curve of system demonstrated that the proposed control algorithm validated its better dynamic and static control quality and strong robustness. The research results show that it is feasible and reasonable for fusion control algorithm, and owns strong robustness.

Key words:Complex correlation system, Uncertainty, Fusion control strategy, HSIC, Robustness

1.Introduction

Up to now the conventional control, such as PID and its improved controller, still occupies the leading position in the field of industrial automation because of being clear in physical concept, convenient in application, and easy in adjustment, but it is unable to do as well as one would wish in the face of system control for complex correlation with uncertainty. The conventional PID is the control of quantitative canonical form control based on strict mathematical model, and it leads to be difficult to construct the mathematical model for complex correlation characteristic between uncertainty and system variables. Even if for the improved PID controller such as the fuzzy controller, theoretically it can describe the qualitative and quantitative control model of application system to actualize the control and be suitable for arbitrary complex object control, but it is only limited to realize the simple control of system in SISO and MISO in the specific actualizing process. With the increase of variables, its inference will get more complex, and if it takes the uncertainty factor into account, then the control will get more difficult to actualize. Although along with the development of intelligence control, there appeared the theory such as expert system, neural control, and genetic algorithm etc and the technology such as adaptive control, self-organizing, and self learning control and so on, there are still many problems facing the control of complex correlation system with uncertainty. This is related to the cybernetics characteristic of system, and therefore it is necessary to explore the robustness of new control strategy.

2.Challenge and cybernetics characteristic

The facing challenge of control is closely related to cybernetics characteristic of system, and therefore it must firstly discuss the characteristic of complex correlation system with uncertainty.

① To be not high in structured degree for controlled complex object, it presents the semi-structured and unstructured feature, and therefore it is difficult to describe by mathematical method.

② To be complex in relation, there is the correlation coupling among variable that reflects system state, so it is hard to realize the decoupling of control variables and simple control of single variable.

③ To be serious nonlinear characteristics in controlled object, it is very difficult to make the quantization processing.

④ To be scattered, time varying, unknown and random in system parameter, it owns the unknown and time-varying feature of time lag.

⑤ To complex in system environment, the interfering of external environment is always unknown, diverse and random, and it owns uncertainty feature. In view of being difficult to build the mathematical model for this kind of complex controlled object, therefore it adopts generalized knowledge model to describe the system control model because of being difficult to realize the control of complex correlation system with uncertainty by conventional PID control and modern control method based on state space description, owing to be strict control based on mathematical model for the above mentioned control method[1-2].

The challenge faced by complex system control mainly represents the following aspects. The conventional control is the control based on deterministic math model, but the uncertainty leads to be difficult to build the math model or the parameter and model structure change in a big range, some industrial interfering is difficult to make the forecast, and its control process belongs to the ill structure, and therefore it is hard to realize the expected control by conventional control. The input and output of conventional control system is difficult to make information interaction for external environment, and it does not accept the information data of non-quantitative form, but the intelligent control can accept information data of non-quantitative form as the input and output. The task of conventional control is unitary, and its output maybe is the constant value (regulating system) or follows the desired motion trajectory (track system), but the task of complex system is complicated. It often requires automatic planning and decision-making ability, and therefore the conventional control can not meet the control requirements. The mature theory and technology can be used for reference for linear control problems in conventional control problems, but it is lack of effective solution methods for highly nonlinear control problems. The conventional control does not own the ability fused the knowledge such as controlled object, environment and human being control strategy etc, and its application scope is limited to the simple system control. The control pattern of conventional control is correspondingly simpler, but it is difficult to represent the characteristic of hybrid control process and the generalized control model based on knowledge by mathematical description form, and to embed other control strategy. For example, it can not adopt the multi-modal control pattern that can make the mutual combination among the qualitative control and quantitative control as well as closed loop control and open loop control etc. The structure of conventional control is fixed, and it does not own the abilities such as variable structure，self-optimizing, self-learning, self-adaptive and self-organization etc, and of course it does not own any abilities such as self-compensating, self-reparing and judgment and decision making.

Facing the above challenge, by means of intelligent control strategy it can obtain more satisfactory solution. For correspondingly complex system, it generally adopts the input-output description method for system, namely it adopts black-box method to analyze the dynamic and static characteristic of control system, and it introduced lots of human being control experience, control expert knowledge, intelligence and skill of field operator etc. The controller design is based on generalized control model, and it combines mathematical model and knowledge system. It adopts the control strategy fused human being intelligence, and obviously the conventional control strategy is incapable of action for this[3-4].

3.HSIC based control strategy

The intelligence of human being mainly reflected in the following aspects of observing, learning, understanding and cognizing the thing ability, and namely it understands and adapts the abilities of various behaviours including the ability of control behaviour. The intelligent control is a sort of automatic control technology, and the emphasized precondition is that it is necessary to carry through artificial interference in actualizing control process, and it can autonomously drive the intelligent controller to achieve expected control objective. The HSIC based control strategy mainly reflected in simulating human being intelligence, the control algorithm is intuitionistic, and the solution of seeking solving problem is from human being itself. When the control strategy simulates the control behaviour of human being, it can enter on two aspects of control function and control structure. Based on artificial intelligence and automatic control theory, it can summarize up the cognition for controlled object characteristic of control expert and actual operator and its experience, and by means of production rule it describes its inspiring, intuitive inference and control behaviour. Based on characteristics memory and online character recognition, it can abstract the dynamic characteristic model of system, with the help of control strategy combined with qualitative decision, quantitative control and open-loop and closed loop control, it can actualize multi-model control. In the aspects of function structure, it is different to information process and decision organization of different layers. In view of being measurable in system error and its change rate, it can make the characteristics memory and online character recognition for control system, and thereby it can construct the control model and control algorithm based HSIC[5-8]. The generalized control model is shown in Figure 1.

Figure 1.Generalized control model based on HSIC

3.1.Static characteristic of HSIC

In the control of HSIC, the static characteristic of controller is shown in Figure 2. In the Figure 2,eshows the system error, andupresents the output of HSIC controller.

In the Figure 2, for convenience, in the followingec shows the error change rate, the motion trajectory ofOA-AB-BC-CD-DE-EF-FH-HG-GIrepresents the transfer process of system state in system motion space, and at the same time it also displays and demonstrates the adopted control pattern in the adjusting control process. For example, it adopts proportional control pattern in sectionOA, and the output isu=Kpe. In which,Kpis a proportional coefficient, and its feature ise×ec>0. Aftere reaches the first maximum valueen1, it comes into the restrained control pattern, and therefore its operation range is over interval [0,en1]. The feature of sectionABis that the original proportion coefficientKPis changed asKPk, and in which,k≯ 1, namely the sectionABis a restraining control section. When it reaches pointB, the output is reduced tou01=kKpen1. After that, it comes into the hold control pattern ofBCsection. The feature of sectionBCis that the output of controller keeps no change, but the errore is gradually reduced to zero from the error maximum value. And after sectionAB, the motion trajectory ofCD-DE-EFandOA-AB-BCcan make the likeness analysis, but the action direction of current period is inversely to the previous period because of its system errore being negative. Similarly, it can analyze the motion trajectory change of sectionFH-HG-GIfor the third period, and after through many periods, the system can finally reaches the expected steady state.

Figure 2.HSIC static characteristic

3.2.Dynamic characteristic of HSIC

The dynamic characteristic of system can be expressed by response characteristic of system output, and the dynamic response characteristic of human simulated intelligent controller is shown in Figure 3.

From the analysis of response curveOA-AB-BC-CD-DE-EF-FG-GHof outputyit can be seen that there is a certain relation between system errore and its change rateec in different output time section. From relation analysis amonge，ecandy, it can find corresponding control strategy, that is to say that it can adopt different control strategy in terms of different error characteristic pattern of system. For example, inOAsection (e>0,ec<0),ABsection (e<0,ec<0),BCsection (e<0,ec>0),CDsection (e>0,ec>0) andDEsection (e>0,ec<0) and so on, it can respectively adopt different control pattern aimed at different situation. For instance, ife×ec>0∪e=0 ∩ec≠0 then it can adopt the proportional control, and the control is changed with error proportion. Ife×ec<0∪ec=0, then it can adopt hold control, and the controlu holds the accumulated sum of errore extremum. Ife×ec>0∪e=0 ∩ec≠0, then it takes proportion control, and the controlu changes with error proportion. Ife×ec<0∪ec=0, then it takes the hold pattern, and the previous hold value should be greater than the accumulated sum of errore extremum. Ife×ec>0∪e=0 ∩ec≠ 0, then it still adopts the proportional control pattern, and so on. After through repeatedly many times control of proportion-hold pattern, finally the convergence value of controlu is a constant value, and the convergence value of system errore goes to zero.

From analysis of the above static and dynamic characteristic, it can be seen that the basic control algorithm of HSIC is to simulate the thinking process of human being. The control essence is that aimed at different system error feature, it adopt different control pattern and control algorithm.

Figure 3.Dynamic characteristic of HSIC

4.Strategy and algorithm of fusion control

The fusion control strategy is based on the generalized control model, and for simplifying the control system structure, it can be fused into the strategy and algorithm such as control expert knowledge, control experience of human being, wisdom and skills of field operator as well specific control rule, inference organization and knowledge base etc, and forms the fusion controller based on human simulated intelligence. The outstanding advantage of fusion control model is as the following. It can conveniently build the knowledge set and control rule base by production rule “If 〈condition〉．Then 〈action〉”, and it is better in modularity and naturality because of being no direct relation among rules. Aimed at each piece control rule under different condition, it can independently make addition, deletion and modification, and it owns good ability to adapt the change of environment. Based on the basic characteristics of combination open loop and closed loop control, the strategy can be divided into two kinds of system error dynamic characteristics pattern. ① Ife·ec≤0 ore=ec=0, then it adopts the half-open loop hold control pattern. ② Ife·ec≥0 ande+ec≠0, then it takes proportional control pattern. The original control algorithm of HSIC is as follows.

In the original control algorithm,eis the system error,em,jisjthpeak value of system,ecis the error change rate,kis the constraint coefficient,KPthe proportional coefficient, andUis the output of controller. The feature of original control algorithm is to actualize the double pattern control, that is to say that for different control pattern it adopts different control strategy, and it realize the control for controlled object by means of the alternate control form of open loop and closed loop. Whene·ec>0∪e=0ec≠0, it takes proportional control pattern, and whene·ec<0∪ec≠0, it adopts hold control patter. After being fused such as adjusting skill of human being, practice experience, expert knowledge, skills and wisdom of operating, the specific control algorithm can adopt the structured English description method to make induction, and for example, it can construct the following control algorithm aimed at the complex correlation system with uncertainty.

Ife·ec≥0 andec≠0 Then

The above algorithm fused the human being intelligence, and it simulated the thinking process of human being. Its outstanding advantage is that it is unnecessary to own more experience knowledge for designer, and the control system is strong in robustness, fast in response speed and sensitive in error change, and therefore it is very high in control precision.

5.Simulation and its analysis

The actual complex correlation system with uncertainty is various, and the complicated degree is also different. The type and degree of uncertainty are difficult to make measure, and it does not a determined measure standard, and therefore the actual control is very difficult to make simulation because of needed building simulation model for system simulation. Generally it can select a more representative math model, through changing the order and parameter of system model it can be realized by means of validating strong robustness of control algorithm. If the robustness is very strong then the control strategy is advisable. Without loss of generality, it takes object model of two-order with time lag as an example, and it takes strong pulse signal to replace the impact of order and parameter of system model as well as external interference, and examines the response characteristic for control strategy and control algorithm. Compared with PID control algorithm, if it is to excel the PID control in robustness performance, adjusting time, rise time, overshoot performance and steady control precision, the control strategy is advisable. Assume the model of controlled object as follows.

In which, the model parameter takes asτ=10 s,K0=4.134,T1=1.0 s,T2=2.0 s. For convenience to compare the control effect, it respectively adopts intelligent fusion control algorithm (HSIC) and PID control algorithm to conduct the control for controlled object of two-order with time lag. Under the condition of strong pulse interference, it examines the response characteristic of process. Here assume that whent= 15 s, it joins an external pulse interference signal, and its pulse amplitude is respectively as 0.5 and 1.0, the pulse width is 0.2 s. Figure 4 and Figure 5 are the response contrast curve of two sorts of control algorithm under the condition of different strong interference.

Figure 4.Response under external interference

Figure 5.Response under strong pulse interference

From the contrast of response curve with PID control algorithm, it can be seen that the fusion control strategy (HSIC) represents very strong robustness, and under the action of strong pulse interference, the response curve of fusion control is hardly changeable. It demonstrated that for the response of HSIC, it is fast in rise time, short in transition process, smooth and steady in response curve, and high in steady control precision. The strategy of HSIC must be conducted to adjust for control process in each control period in terms of error size, change direction and its parameter change trend, and it forces the error to go into zero. The fusion control strategy based on HSIC belongs to the nonlinearity control, and it allows the control amplitude to be larger, and therefore the Bang-Bang control strategy can be fused into fusion control strategy so as to improve the process response speed and shorten the transition process time. In addition, it can also fuse other control rule to make the large amplitude overshoot be restrained so as to avoid producing oscillating resulted in excessive adjusting of introduced Bang-Bang control.

6.Conclusion

From the above simulation it can be seen that based on the strong robustness of adaptive fusion control strategy based HSIC, it is unnecessary to build strict mathematical model, and it is less demand of experienced knowledge for designer, and therefore it is a more wise choice for the control of complex correlation system with uncertainty. The above simulation result under the condition of strong interference validated that the fusion control strategy based on HSIC is feasible and effective, and it owns very strong robustness.

References

[1]YANG Zhi, LI Taifu, SHENG Chaoqiang et al. Complex Correlation System Control Based on Human Simulation Intelligence[J]. Journal of Chongqing University, 2002, 25(7): 9-11.

[2]LI Taifu, FENG Guoliang, ZHONG Bingxiang, et al. Analysis on Control Strategy for a Kind of Uncertain Complexity System[J]. Journal of Chongqing University, 2003, 26(1): 4-7.

[3]LI Shiyong. Fuzzy Control Neurocontrol and Intelligent Cybernetics[M]. Harbin: Harbin Industry University Press, 2002.

[4]YI Jikai. Intelligent control Technology[M]. Beijing: Beijing Industry University Press, 2004.

[5]LI Zushu, TU Yaqing. Human Simulated Intelligent Controller[M]. Beijing: National Defense Industry Press, 2003.

[6]XIONG Renquan, QIAO Zhenghong. Control strategy of supply system based on HSIC[J]. Journal of Sichuan ordnance, 2012, 33(1): 76-78.

[7]CHEN Wenbai, GAO Shijie, WU Xibao. HMCD-based dynamic motion control strategy of humanoid robot on a horizontal bar[J]. CAAI Transactions on Intelligent Systems, 2012, 7(6): 501-505.

[8]ZHANG Cuiying, TIAN Jianyan. Multi-modal Control of Temperature in the Reheating Furnace Based on Human-simulated Intelligent Control Theory[J]. JOURNAL OF TAIYUAN UN IVERSITY OF SCIENCE AND TECHNOLOGY, 2008, 29(1): 12-15.

基于仿人智能融合控制算法的鲁棒性仿真研究*

朱顺兰† 1，黄蕙1，张四平2

1重庆工业职业技术学院 自动化学院, 重庆401120；2中冶赛迪(集团)电气技术有限公司，重庆400013

摘要：为了克服控制算法因鲁棒性差难以实现对不确定性复杂关联系统控制的缺陷，探讨了仿人智能融合控制算法的鲁棒性。讨论了传统控制面临的挑战，粗略地分析了不确定性复杂关联系统的控制论特性，研究了系统的仿人智能融合控制策略，针对具体对象构造了相应的控制算法。以二阶时滞对象控制为例，分别采用PID与仿人智能融合控制算法进行了仿真对比研究，实验的系统响应曲线验证了该算法具有良好动静态控制品质与鲁棒性能。研究结果表明：提出的融合控制算法是可行与合理的，具有很强的鲁棒性。

关键词：复杂关联系统；不确定性；融合控制策略；仿人智能控制；鲁棒性

中图分类号：TP273

DOI:10.3969/j.issn.1001-3881.2014.06.021

Received: 2013-11-05

*Project supported by Chongqing Education Commission (2012-09-3-314)

† Shun-lan ZHU, E-mail: 756786459@qq.com