Behaviour of intelligent autonomous robotic IAR



Behaviour of Intelligent Autonomous Robotic IAR

OUARDA HACHOUR(*) AND Nikos Mastorakis(**)

(*) Robotics and Artificial Intelligence Laboratory

Development Center of Advanced Technologies Baba Hassen Haoulh Oukil

Algiers Algeria

Phone: 213 (021) 35-10-18,35-10-40,35-10-75. Fax: 213 (021) 35-10-39

(**) Hellenic Naval Academy

Department of Electrical Engineering and Computer Science

Terma Hatzikyriakou, 18539

Piraeus, Greece

Abstract: - A robotic system capable of some degree of self sufficiency is the primary goal of IAR The focus is on the ability to react and on being self-sufficient to evolve and carry out tasks in various environments by themselves. To deal with Artificial intelligence, the robot must then be able to understand the structure of this environment to find a way towards its target without collisions. So, data processing, pattern recognition, reasoning learning, interpreting and decision–making are the most suitable factors to be endowed with the robot. Thus, IAR designers search to create intelligent system to take decision like human in real time. This behaviour near to the human one in recognition, learning, decision-making and action is captured to deal with autonomy requirements such as thermal, energy, and mechanical design and communication management to present a real intelligent task. The ability to acquire these faculties, to treat and transmit knowledge constitutes the key of a certain kind of intelligence. This kind of intelligence represents the human ambition in the construction and development of an intelligent machine, particularly, the autonomous mobile robots which are the aim of all intelligent tasks. Afterwards, a survey of intelligent sensor strategies, emergent techniques for intelligent control systems, coordinates behaviour and distributed control architectures, and massively parallel computers are presented. Finally, we present a brief survey of IAR applications.

Key-Words: -intelligent Autonomous Robotic (IAR), Intelligence, Multi-Agent systems, Distributed Control, Soft computing, Autonomy Requirements.

1 Introduction

Robotic systems capable of some degree of self-sufficiency are required in many fields. Thus , IAR designers search to create suitable architecture to achieve intelligent behaviours like human in real time. In fact, the application of these systems to industry, underwater, planetary explorations and high way is in increasing importance for several reasons as : reducing health hazards and minimizing human operator fatigue. The IAR must be able to perform intelligent behaviours in real time. Through sensors they obtain information and act upon the world through actuators. The sensor may be connected to actuators by a simple signal processing or it may involve goal interpretation obstacles avoidance and other aspects of reasoning . In this paper, we present the behaviour of IAR. The aim of this work is to give a real intelligent task to deal with autonomous requirements and intelligent component which ensure reactivity and intelligent autonomous to these systems. Afterwards, the autonomy requirements and

intelligent components are discussed. Then, a brief survey of IAR applications is presented.

2 The performance of IAR

Nowadays, IAR can carry out task in various environments by themselves like human. These intelligent autonomous systems are very useful in many fields as industry and planetary explorations. However, they are all semi-autonomous and need some human operators. In fact, The future factory need all flexible and robust IAR. Additionally, IAR are becoming more and more interesting for underwater, terrestrial, and space applications. These mechanical systems are constructed to respond any traditional working as in construction and agriculture. Previously, certain industrial operations required human skills may be tedious and exceptionally hardly ever. Above all, repetitive operations can result in reductions in quality control, as in visual inspections tasks. Also, these repetitive actions may be hazardous health risks as exposure to unsafe materials like radioactive and high pressure in underwater applications. So, the presence of human workers in these environments may be perilous which need the necessity to be replaced by intelligent systems, these systems can move, react, and carry out tasks in various environments by themselves like human [24]. These untethered systems become then free of constraints than is currently used with remotely operated systems [11]. This involves that intelligence and sufficient power must be integrated at a higher level while the robot communicate with the environment. So, IAR must be able to perform purposeful behaviours in the real-world. These vehicles capable of coping with real dynamic worlds populated by other robots and humans constitute powerful test-beds to study and exercise the multi-faceted aspects of intelligent [12].

3 Necessity of Intelligent Autonomous Robot in terrestrial and planetary interaction

The theory and practice of IAR are currently among the most intensively studied and promising areas in computer science and engineering which will certainly play a primary goal role in future. These theories and applications provide a source linking all fields in which intelligent control plays a dominant role. Cognition, perception, action, and learning are essential components of such-systems and their use is tending extensively towards challenging applications (service robots, micro-robots, bio-robots, guard robots, warehousing robots). Many traditional working machines already used e.g., in agriculture or construction mining are going through changes to become remotely operated or even autonomous. Autonomous driving in certain conditions is then a realistic target in the near future. Technology has made this feasible by using advanced computer control systems. Also, the automotive industry has put much effort in developing perception and control systems to make the vehicle more safe and easier to operate. To perform all tasks in different environments, the vehicle must be characterized by more sever limits regarding mass volume, power consumption, autonomous reactions capabilities and design complexity. Particularly, for planetary operations sever constraints arise from available energy and data transmission capacities, e.g., the vehicles are usually designed as autonomous units with: data transfer via radio modems to rely stations ( satellite in orbit or fixed surface stations) and power from solar arrays, batteries or radio-isotope thermo electric generators (for larger vehicles).

1. Autonomy Requirements

(Thermal, Energy, Communication Management, and Mechanical design)

Several autonomy requirements must be satisfied to well perform the tasks of IAR, this is summarized in some in the following section.

1. Thermal

To carry out tasks in various environments as in space applications, the thermal design must be taken into account, especially when the temperature can vary significantly. At ambient temperatures, the limited temperature -sensitive electronic equipment on-board must be placed in a thermally insulated compartments [23] . The thermal environment of Mars challenges the thermal control system. In the course of a Martian day the temperature can vary from 140K to 300K. As the mechanical and electro-mechanical components should operate without thermal regulation at ambient temperatures, the limited temperature –sensitive electronic equipment on-board like scientific experiments, sensors cameras, power, and data handling interface must be placed in a thermally insulted compartments. All cables exiting this compartment must be especially insulated.

3 Energy

For a specified period, IAR can operate autonomously, one very limited resource for underwater and space applications are energy. So, IAR usually carry a rechargeable energy system, appropriately sized batteries on-board.

5 Communication Management

The components on-board the vehicle and on-board the surface station must be interconnected by a two-way communication link. As in both underwater and space applications, a data management system is usually necessary to transfer data from IAR to terrestrial storage and processing stations by two- way communication link. Indeed, the data management system must be split between components of the vehicle and surface station. Thus, the vehicle must be more autonomous and intelligent to perform and achieve the tasks. Due to the limited resources and weight constraints, major data processing and storage capacities must be on the surface station. Although individual vehicles may have wildly different external appearances, different mechanisms of locomotion, and different missions or goals, many of the underlying computational issues involved are related to sensing and sensor modelling spatial data representation, and reasoning.

3.1.4 Mechanical design

The mechanical design of IAR is the result of an integration approach considering several criteria related with perception, control, and planning issues in addition to structural design and other mechanical requirements [20].

3.2 Criteria to satisfy by vehicles to be autonomous and intelligent

To evaluate the performance of IAR which are intelligent and autonomous vehicle, the robot must perform the following criteria:

3.2.1 Intelligence

The intelligence is the capability of IAR to achieve missions in various environments by themselves like human. The use of Expert Systems (ES), Fuzzy Logic (FL) and Genetic Algorithms (GA) is necessary to bring the behavior of Intelligent Autonomous Vehicles (IAV) near to the human one in recognition, learning, decision - making and action. These faculties must make the robot able to achieve these tasks: to make ones way towards its target, and to avoid obstacles capturing the behavior of a human expert.

2. Navigation

Navigation is the ability to move and on being self-sufficient. The IAR must be able to achieve these tasks: to avoid obstacles, and to make one way towards their target. In fact, recognition, learning, decision-making, and action constitute principal problem of the navigation.

2. Autonomy, Safety, and Reliability

A robotic system capable of some degree of self-sufficiency is the primary goal of IAR. Thus, the robot must achieve his task with more autonomy and intelligence. Also, the vehicle react to unknown static and dynamic obstacles with safety not to endanger itself or other objects in the environment[15]. Near Safety, the reliability is taken into account in the field of robotics; it is the probability that the required function is executed without failure during a certain duration. Reliability depends on the system design, the used materials and components, and the development process and manufacturing process.

4 Robot and sensors

To detect all possible obstacles, the robot is supposed to have vision system (camera). To operate in certain dynamic environment, the use of two or more sensors can guarantee to deliver acceptably accurate information all of the time. Thus, the redundancy can be useful for autonomous robotic sensory systems as in the human sensory system [2], multi sensor (sensor data fusion) integration has them received a good deal of attention in recent years, it is used as a technique useful precise and achieve a specific task and obstacle avoidance.

5 Mapping and cognition of the environment

The use of map to structure our environment has often been more efficient than previous techniques. The difficulty in building a map of the environment lies in the cognition representation. For same types of navigation, it is more advantageous to use an implicit one. In the intelligent robot behavior, this environment model map has an important role to play. So, building a map of the sensory input space is more interesting especially when the external environment is unknown. In this context, to structure our environment we give this subdivision the following points:

1. Geometrical level

This task consists on the "Cartographer work”. The flat earth model describes the viewable area of the world as an ideal planner surface S defined by the points of contact of the objects projected onto the ground. Considering the point of map S (x, y) down on the surface S. This point is the intersection of the line

L (x, y) with the surface S, this is how the projection is given onto the flat earth model and shadow projections.

2. Topological level

This one illustrates the features map; it consists of the decomposition on free and occupied space, and gives a relationship between the free spaces. In this case, a grid-map is more suitable for the unknown environment. In addition to determine a valid, safe path grid-mapped environments, an intersection of parallel horizontal and vertical lines is the key of digital field analysis, that is the way to interpret safety and danger areas. We mention below this organization, the obtained grids are homogenous in dimensions.

3. Semantic level

In the previous step, the topological level provides fine entering of organization map and offers all features

associated with the appropriate environment. In this context, the homogenous grid-map obtained must be associated with the appropriate label useful in real-time, learning, generalizing, and approaching human-like reasoning for each particular situation, this association is achieved by:

5.4 Path planning

The multi-level structure of path planning and execution propounded in provides a basic framework for dealing with problems in the control of autonomous vehicles. There are three basic levels of path planning and execution.

1. The planner

A global path planner uses priori knowledge (a map) to plan a plausible road.

2. The navigator

A local path planner, uses the plan of the planner a guide, but provides more precise routing according to obtained terrain information locally.

3. The pilot

Execution of simple vehicle movement routines.

5 combination of technologies GA, FL, ES, NN for intelligent control systems

Today, Researchers have at their disposal, a computational tools such as FL and GA that are more effective in the design and development of IAR than the predicate logic based methods of traditional Artificial Intelligence. Genetic Algorithms GA, Fuzzy Logic FL, Expert system ES, and Neural Network NN are well established as useful technologies that complement each other in powerful hybrid system. Hybrid intelligent systems are now part of the repertoire of computer systems developers and important research mechanisms in the study of Artificial Intelligent. The first and most advanced integration of intelligent technologies is the hybrid GA and FL is to synthesis the capability of ES to capture expert domain knowledge in a inference- based system. The integration of NN and ES has proven to be a way to develop useful real-world applications, and hybrid systems involving robust adaptive control. Other new hybrid systems introducing adaptative resonance theory and Chaos theory emerged. In particular, the fuzzy ARTMAP, is developed by Steve Grossberg and Gail Carpenter[5] to model the living mind and the concept of symbolic dynamics by Chaos NN and adaptive system related Chaotic dynamics to intelligent computations.

5.1 Robust Adaptive Control (for non-linear and linear plants)

The ideas and principles behind design simulation and implementation of robust adaptive control systems for identification and control are oriented towards recently developed adaptive design tools which are combined into recursive design procedures for non-linear systems with unknown parameters and unmeasured states. Such controllers possess stability, convergence, and transient performance properties superior to those of some traditional adaptive schemes. Their applications include electronic braking adaptive cruise control, active suspension, heavy-duty vehicles and jet engines.

5.2 Expert System ES: is a computer program that human experts, in a narrow domain, dealing with specialised knowledge, generally possess function. Es is able to draw conclusions without seeing all possible information in an efficient manner [18,19].

5.3 Genetic Algorithms GA

The genetic algorithms, which are evolutionary, have recently emerged from study of the evolution mechanisms and are searching strategies suitable for finding the globally optimal solution in a large parameter space; they are based on learning mechanism. GA’s has been theoretically and empirically proven to provide robust search capabilities in complex spaces offering a valid approach to problems requiring efficient and effective searching [17,26]. Before the GA search starts, candidates of solution are represented as binary bit strings and are prepared. This is called a population. A candidate is called a chromosome [7,8]. Also, an evolution function, called fitness function, needs to be defined for a problem to be solved in order to evaluate chromosome. This evaluation is the goal of the GA search and goes as follows : two (02) chromosomes are chosen randomly from a population are mated and they go through operations like the crossover to yield better chromosomes for next generations. Because the population size is fixed, the offspring produced by genetic manipulation process are the next population to be evaluated. We must specify a stopping criterion to determine execution of the GA. This search mechanism is associated with mutation a mechanism of probabilistic to change a bit 0 to 1 or vice versa (see Fig.1). The GA treats mutation only as a secondary crucial operator after crossover. The crossover is the comparison operator; this operation takes chromosomes and swaps part of their genetic information to produce a new chromosome (see Fig.2).

Fig. 1 : Example of Crossover on single point

Fig. 2 : Example of Mutation in the second bit.

4. Fuzzy Logic FL

To build machines that are able to perform complex task requiring massively parallel computations, knowledge the environment structure and interacting with it involves abstract appreciation of natural concepts related to the proximity, degree of danger, etc. The implied natural language is represented through fuzzy sets involving classes with gradually varying transition boundaries. In effect, human reasoning is not based on the classical two-valued logic because this process involves fuzzy thrusts, fuzzy deduction rules, etc. This is the reason why FL is closer to human thinking and natural language than classical logic [19,29,30,31]. Also, FL can be viewed as an attempt to bring together conventional precise mathematics and human-like decision-making concepts [4]. FL can be a valid approach solving control problem in a wide range of applications. In particular, embedded architectures are likely to use fuzzy logic in the future for dedicated applications. The Fuzzy Logic Controller FLC can be considered as a system given an input vector computes an output vector by a linguistic rule (as an example: see Fig.3). To define the complexity of a fuzzy controller we consider some typical parameters such as the number Input and the number output, the dimension of the rule base, the number of membership functions per input, the precision and the methods chosen to performed the three well known steps: fuzzification, inference and defuzzification steps. As an example, D=(, B= (, D The parameters: (,(, and D are illustrated in the Fig.4. D is an intermediate distance between P1 and Pi, Pi and Pg, and P1 and Pg. The distance between P1 and Pi for example is given by: [pic] (1). Where: the points P1 (X1,Y1), Pi (Xi,Yi) and Pg(Xg,Yg) are the co-ordinate of respectively to current point, intermediate point and visual point (we calculate point to point until the visual point becomes the target one). The vehicle going to a given target avoiding a static obstacle is shown in this figure. At first, from a given detected position Pi (Pixel by pixel), and from a given visual position Pg, the vehicle gets absolute obstacle position Pi and Pg from current point P1 to calculate the angles ( and (. Then, after updating its position (P1 becomes Pg) and the target position (Pg becomes the target one) based on its Cartesian coordinates, the vehicle must avoid the obstacle to get the target position. Afterwards, the vehicle recognizes the static danger degree ( and safety degree ( between itself and the obstacle using a fuzzy reasoning and inference. Second, using ( and ( the vehicle decides the avoidance direction ( by a decision table written with productions rules and then avoidance direction vector . The main problem of this approach is that does not encounter several obstacles at the same time and does not take into account the obstacles sizes. The static danger degree ( and safety degree ( are given by :

[pic]

[pic]

Fig.3: Fuzzy Model.

Fig.4: Vehicle obstacle avoidance mode

5.4.1 Fuzzification

Fuzzy logic is based on the concepts of linguistic variables and fuzzy sets. A fuzzy set in a universe of discourse U is characterized by a membership function mf which assumes values in the interval [0,1] . A fuzzy set F is represented as a set of ordered pairs, each made up of a generic u which is in U and its degree of membership mf(u). A linguistic variable x in a Universe of Discourse U is characterized by a set W(x) = (Wx1,…,W xn) and a set M(x) = (M x1,…,M xn) where W(x) is the term-set i.e., the set of names the linguistic variable x can assume, and Wxi is a fuzzy set whose membership function is Mxi. Each element of which is associated with a membership function. The membership functions of the fuzzy sets can be trapezoidal, triangular, and sinusoidal sets. As an example, the Fig.5 represents the triangular membership functions:

Fig. 5 : Membership functions of distance D

Where the membership labels for distance D are defined as Near (N), Medium (M) , and Far (F). D is an intermediate distance between the points X1 and Xi, Xi and Xg, and X1 and Xg (N, M, F are linguistic variables treated in this example)

5.4.2 Inference

The rules governing a fuzzy system are often written using linguistic expressions, which formalize the empirical rules by means of which a human operator is able to describe the process in question using his own experience. More, It is a way of linking input linguistic variables to output ones. If x and y are taken to be two linguistic variables, fuzzy logic allows these variables to be related by means of fuzzy conditional rules of the following type

IF (x is A ) THEN (y is B)

Where (x is A) is the premise of the rule, while (y is B) is the conclusion. The premise defines the conditions in which the conclusions define the actions to be taken when the conditions of the premise are satisfied. More specially, the degree of membership of the premise is calculated and through application of a fuzzy logic inference method to the conclusion, it allows the output y to be determined. In general in a fuzzy conditional rule “if premise then conclusion” is made up of a statement in which fuzzy predicates Pj of the general form ( Xj is Aj) are combined by different operators such as the fuzzy operators AND and OR .In this case Xj is a linguistic variable defined in the Universe of the discourse and Aj is one of the names of the term set of Xj. The following is an example of fuzzy conditional rule using operators: IF (P1 AND P2) THEN P4. Where P1 = (X1 is A1 ), P2 = (X2 is A2), P4 = ( Y4 is B4 ). To apply an inference method to the conclusion, it is first necessary to access the degree of membership of the premise, trough assessment of the degree of membership of each predicate Pj in the premise.

5.4.3 Defuzzification

Defuzzification is the output of the fuzzy system, it is a decision-making logic ( written in a formula) adopted for the compute of the real value of the output. The final decision (defuzzification) is achieved to give the output of fuzzy controls and to converts the fuzzy output value produced by rules. The system must decide how to give the right output using FL from a fuzzy linguistic formulation. A defuzzification stage generally performed using the centre of gravity method, MAX-MIN method, …etc. This step will give The real value output , as an example, a defuzzification formula of centre of gravity is given by

G = (the sum of (ui * gi ) / the sum of (ui) )

1 ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download