Robot Motion and Control 2009

This article is concerned with methods of computing a path in 3-DOF space that describes the desired translational motion of an aerial vehicle. Trajectory planning is an optimization problem which generates an optimal trajectory between two configurations in the state space, considering a given performance index (time, energy or distance). Its feasibility depends on the choice of the optimization method, the performance index and a number of constraints from various nature, the latter depending essentially on the vehicle itself (architecture, dynamics and actuation modes) and the environment in which the vehicle moves (endurance, airspeed, altitude, landing and takeoff modes, etc.) [3, 5, 12, 13]. The classical differential geometry curve theory is a study of 3D space curves with orthogonal coordinate systems attached to moving points on the space curve [1, 2, 4, 11]. The atmosphere is considered to be an isotropic and homogeneous medium, i.e. when there is no wind and the air density is constant with altitude. Classically, in motion planning and generation, methods such as continuous optimization and discrete search are sought. In [9] randomized motion planning algorithm by employing obstacle free guidance system as local planners in a probabilistic road-map framework are presented. In [13], path planning of autonomous fixed wing aircraft is based on a learning real-time A* search algorithm, considering only the motion on a horizontal plane. A family of trim trajectories in level flight is used in all these references to construct paths. In [7], plans are described as the concatenation of a number of well defined motion primitives selected from a finite library.

In the synthesis of automatic controllers of an unmanned aerial vehicle (UAV) as for other dynamic moving vehicles a mathematical model of motion should be used. Advanced utilization of the vehicle dynamics is only possible if a control system is designed considering as exact as possible characteristics of the vehicle. The proper design of the controller is critical to the safety and before in-flight tests safe operation and landing must be considered.

Identification of the model of an aircraft dynamics is the most reliable method to obtain a model of motion. Theoretical analysis and laboratory (wind tunnel) tests are costly and do not include effects of some phenomena occurring in real flight. However, an in-flight experiment with only little known object is very risky.

In the recent years, the problem of finite-time convergence for the continuous dynamic systems has attracted more attention of the researchers [1, 4, 2]. On one hand, accomplishing the stabilization control task in a finite time seems to be more natural from a practical point of view than the infinite-time solution characteristic for the

classical

asymptotic results. On the other hand, as presented in [1], the finite-time stable systems are more robust, guaranteeing improved rejection of the non-vanishing (persistent) low-level disturbances. Moreover, the continuous finite-time stabilizers make it possible to avoid the so-called chattering phenomenon intrinsic for example for low-order sliding mode controllers. Utilization of the finite-time stability results in the control design for mobile robots can have an additional merit coming from the possibility of assessing the settling time interval in the closed-loop system. It can facilitate the motion planning stage.

Introduction The mechanical nonholonomic systems attract research attention in the field of automation and robotics. It results from commonness of such systems in the real environment. In spite of this problems that occur during control of those systems encourage continuing searching for new and better solutions. Difficulty with reaching separated points in state-space configuration, when full controllability is available, results from the kinematic structure of those systems.

Scientific interest in mobile robotics has been growing rapidly in recent years, which is due to increasing significance of autonomous vehicles in many applications. In order to ensure a suitable level of robot’s autonomy automatic realization of fundamental motion tasks is required. Hence, the control problem at robot platform kinematic level becomes a very important task.

Cooperative robotics is a currently intensively investigated area of science. Great development in technology has attracted research interest in many countries on this subject. It is caused by availability of reliable, efficient and inexpensive components of mobile robots. In addition, some tasks can be executed more efficiently using multiple simple, low-cost mobile robots instead of one bigger robot. Multi-robot systems have a wide range of applications: service robots, transportation, exploration, mapping and rescue systems, surveillance, security and many others.

On the other hand, control design for multiple robots is much more complex due to a distributed characteristic of the system and its complexity. There are many approaches which can be conventionally partitioned into three groups: virtual structure approach [2, 7], behavioral approach [4, 11], and leader follower scheme [3, 8, 9]. Each of them is more suitable for particular application and less for the others.

Dynamical models of flexible-link robot manipulators are generally described by a set of coupled ordinary and partial differential equations, which gives rise to series of mathematical control problems in infinite dimensional spaces [1, 2, 3, 7]. However, finite dimensional approximate models obtained by the assumed modes and finite elements methods are used more frequently for solving the motion planning and stabilization problems [6]. It should be emphasized that the majority of publications in this area is concentrated on planar manipulator models with a free end. To study spatial manipulators with a tip mass, the mathematical model that describes the motion of a multi-link manipulator under the action of gravity and controls (torques and forces) was proposed in [8].

The controllability property plays a crucial role in mathematical control theory. For linear systems, necessary and sufficient conditions of controllability are given by the Kalman criterion. This criterion also allows to study the local controllability by linear approximation [5]. There is a number of necessary as well as sufficient conditions of controllability for nonlinear systems expressed in terms of properties of the corresponding Lie algebra (see, e.g., [2, 7, 9]). However, in general case, a constructive test of controllability of nonlinear systems is a difficult problem and, to our knowledge, there is no effective estimate of the number of iterated Lie brackets for the test of an appropriate rank condition [1]. In this paper, we use controllability conditions of nonlinear systems based on a modification of the return method (cf. [4, 6, 8]).

A robot’s drive has to exert appropriate driving forces that can keep its arm and end effector at the proper position, velocity and acceleration, and simultaneously has to compensate for the effects of the contact forces arising between the tool and the workpiece depending on the needs of the actual technological operation. Balancing the effects of

a priori

unknown external disturbance forces and the inaccuracies of the available dynamic model of the robot is also important. Technological tasks requiring well prescribed end effector trajectories and contact forces simultaneously are challenging control problems that can be tackled in various manners.

Though the formal mathematical idea of introducing noninteger order derivatives can be traced from the 17th century in a letter by L’Hospital in which he asked Leibniz what the meaning of

D

n

y

if

n

= 1/2 would be in 1695 [1], it was better outlined only in the 19th century [2, 3, 4]. Due to the lack of clear physical interpretation their first applications in physics appeared only later, in the 20th century, in connection with visco-elastic phenomena [5, 6]. The topic later obtained quite general attention [7, 8, 9], and also found new applications in material science [10], analysis of earth-quake signals [11], control of robots [12], and in the description of diffusion [13], etc.

Pneumatic muscles are becoming increasingly attractive robotic drives for those applications in which stiffness of a kinematic chain have to be regulated. The stiffness coefficient of the muscle depends mainly on the level of pressure and its initial tension. These values describe the point of work on the static characteristics of the drive. Such a drive could be treated as a regulated spring. However, from the dynamic point of view the muscle is something more than a simple spring. Generally, three-element models (of the type R, L, C) are used in the literature to present the dynamical effects that characterise the drive [6]. It leads to second order differential equations that describe the features of the single muscle or the whole drive in the neighborhood of the static point of work. Sometimes this simple model is not sufficient and some phenomena of the muscle are modeled by additional inertial elements of the first or second order, and the model is characterized by three or four parameters. Precisely speaking, the muscle has to be treated as an element with distributed parameters and to describe its phenomena, partial differential equations could be used. Unfortunately, such an analytical model does not exist in a general case. A finite element method can be applied to derive a numerical model convenient for specific purposes [10].

Automatic control synthesis for large scale systems is a central capability for creation of next generations of autonomous systems, such as automatic robotic manufacturing systems, and mobile robotics operating in uncharted environments.

Previously, it has been shown that feedback control can be automatically generated for a class of nonlinear systems. In [12] evolutionary algorithms and in [13] Lyapunov functions were used for automatic control synthesis. These methods are not devised to handle switches between operation modes characterized by different dynamic. Nevertheless, mode change is essential for autonomous dynamical systems. On the other hand, considerable progress has recently been made in the direction of discrete abstractions of continuous and hybrid control systems. In e.g. [16] the state space was partitioned by higher dimensional cubes, and controllable system has been shown to be bi-similar to a discrete register.

A motion planning task in robotics and related domains is to steer an object from its initial location

$\pmb{q}_0=\pmb{q}(0)$

to a goal point

$\pmb{q}_f=\pmb{q}(T)$

[3, 7]. The task is particularly difficult when the number of controls is smaller than the dimensionality of a state space. In this paper the class of driftless nonholonomic systems is considered. The systems are described by

$$ \dot{\pmb{q}}=\sum_{i=1}^m \pmb{g}_i(\pmb{q}) u_i= \pmb{G}(\pmb{q}) \pmb{u}, \;\;\;\; \dim \pmb{q}=n > m=\dim \pmb{u}, (13.1) $$

where

$\pmb{q}$

is a configuration,

$\pmb{g}_i$

,

i

 = 1, ...,

m

, are generators (smooth enough vector fields) of system (13.1), and

$\pmb{u}$

denote controls.

If compared with biological systems that routinely exhibit dynamic behaviors in complex environment with surprising adaptivity, energy efficiency and robustness, our robots are still severely suffering from the lack of sensory-motor and learning capabilities [1]. To account for the discrepancy of behavior control in animals and robots, there has been an increasing interest in the study of underactuated robotic systems for rapid, efficient and maneuverable behaviors in the real world.

Walking in rough and unstructured terrain with a legged robot remains to be a challenging task, not only regarding the construction of a robust walking machine, but also looking at the control of such a complex system. Walking robots have the ability to climb over obstacles and can also walk fast in flat terrain. This is possible because the control parameters are adapted to the surrounding terrain. For example, the swing height has to correspond with the obstacle height, elsewise the robot cannot walk over these obstacles. An adaptation of the control parameters according to the environment can be realised in different ways. If there is no environment model present, the robot can only rely on its sensor systems trying to react to influences from the environment. Only the use of an environment model can enable the robot to interact more intelligently with the environment and adapt its parameters in advance. In this way, the robot can avoid many collisions, which speeds up the walking process. In the example, this would mean to increase the swing height and therefore modify the footpoint trajectory with the aim to overstep the obstacle. Because the swing height is only one parameter among many others, it is obvious how expedient an environment model is for a complex walking robot like LAURON.

Over the last two decades, a number of advanced humanoidal bipeds have been developed, e.g. ASIMO, HRP-3, H7, QRIO

TM

, ROBIAN, WABIAN, to name a few of them. It is worth mentioning that progress towards a more human like structure has been made by scientists from the University of Tokyo, who have built a Kotaro robot [2]. Obtaining robot movement similar to that of humans is still a problem due to mechanical construction limits or due to weaknesses of motion generation methods [3]. The research being conducted at Warsaw University of Technology aims at developing biologically inspired motion patterns [11, 10, 9] for legged robots.

Walking machines are a field of intensive research. The wide spectrum of possible applications is the reason why the new prototypes of two-, four-, and more-legged robots are constructed. The engineering of those new machines gives opportunity to optimize and study not only original mechanical solutions but also control algorithms devoted to walking in different environmental conditions with various tasks given to the robot. The term “walking robot” means a machine that performs gait with dedicated limbs [8, 7]. The robot’s performance is based mainly on the continuous information about its state and surrounding obstacles [3]. The basic information about robot state includes the positions and velocities of particular legs links as well as forces exerted on it. For a multi-legged walking machines the mostly considered type of walking is a static gait [6]. More demanding control task is a dynamic walking, which was firstly researched for the case of biped robots [1]. The well-known solutions for machine dynamic gait are presented in [2, 4].

Robot simulators give many opportunities for development of new control algorithms. They are particularly useful for improvement of control strategies by means of learning. Motivation to use a simulator for learning such robot behaviors like gaits in walking robots stems from the fact that the trial-and-error learning on a real robot may be dangerous. Also, the number of simulated runs performed in the given amount of time can be much higher than the number of real experiments. However, any simplification made in a simulator may be exploited by the learning algorithm, resulting in a gait pattern, which cannot be reproduced on the real robot. This “reality gap” problem has been widely discussed in the literature [5, 9].

Mobile robots are still an object of research and are popular as a subject of commercial applications. Some are designed to work in an urban space, where one of the most common mobility barriers are stairs.

The stair climbing task is performed by tracked machines. One of the exemplary solutions is described in [7], where such a robot is equipped with gyroscopes, a tilt sensor, a pair of cameras and a LADAR to support the move on a staircase. Another type of a robot which is able to climb and descent stairs is the hexapod “RHex” described in [8, 12]. The machine combines characteristics of tracked and walking robots for better manoeuvrability. The robots presented in two mentioned examples are unable to move in a horizontal direction, perpendicular to the climbing direction, to avoid unexpected obstacles such as human legs on stairs. Their mobility is insufficient because of the limited number of DOFs available.

The extension of application domains of robotics from factories to human environments leads to implementing proper strategies for close interaction between people and robots. On the one hand, small-scale industrial robots have to learn to get along with human coworkers in factories, and, on the other hand, service robots are a solution for automatizing common daily tasks in domestic environments, due to lack or high cost of human expertise.

The size of an industrial robot, or the necessary autonomous behavior of a service robot, can result in dangerous situations for humans coexisting in the robot operational domain. Therefore, physical issues must be carefully considered, since “natural” or unexpected behaviors of people during interaction with robots can result in injuries, which may be severe, when considering the current mechanical structure of robots available on the market [1].

The presented discussion of the structures of robot control systems is based on the concept of an agent. The agents having physical bodies (e.g., robots) are termed embodied agents. Both the operation of a single agent and the interactions between agents is of interest to us. The operation of an embodied agent is based on the concept of transition functions. The necessary concepts and an example of utilization of those concepts in the specification of a two-robot control system are presented in this paper. The example describes the specification and implementation of a haptic device. Haptic interfaces enable the human operator to acquire a remote sense of touch. The operator interacts with the master manipulator and through it commands the motion of the slave manipulator, but is also able to feel the forces of interaction between the slave and the environment. There are many research problems associated with haptic systems: the construction of the master [1, 4], haptic control [7, 6], stability of the haptic coupling [3]. Our implementation of the haptic system uses robot manipulators. Its formal specification is the subject of this part of the paper.

Recent advances in robotics include the capability of planning a suitable trajectory in order to drive the robot from an initial configuration to a predetermined goal point, or to follow, when possible, a prespecified trajectory even in unknown environments [8]. Various methods can be adopted to accomplish this task. These methods are mainly classified in relation to the capability of the sensors which are employed to map the environment and the obstacles near the robot. When distance sensors and cameras are considered, the trajectory can be planned without colliding with the obstacles (i.e. no force measurements are required), see [3, 12, 13].

Many applications of service robotics are based on object manipulation with multi-fingered mechanical hands, where the fingers should operate in a coordinated fashion to achieve the desired motion of the manipulated object. In the absence of physical interaction between the fingers and the object, simple motion synchronization shall be ensured. On the other hand, the execution of object grasping or manipulation requires controlling also the interaction forces, to ensure grasp stability [1].

An important issue in multi-fingered robotic manipulation is the formulation of the coordinated task. From a purely kinematics point of view, the task is usually assigned in terms of the motion of the fingertips and/or in terms of the desired motion of the manipulated object. In all cases, the next step toward implementation is that of mapping the desired task into the corresponding joint trajectories for the fingers. This operation can be carried out at planning or at control level, but always requires the solution of an inverse kinematics problem.

Robots have recently been foreseen to work side by side and share workspace with humans in assisting them in tasks that include physical human-robot (HR) interaction. The physical contact with human tasks under uncertainty has to be performed in a stable and safe manner [6]. However, current industrial robot manipulators are still very far from HR coexisting environments, because of their unreliable safety, rigidity and heavy structure. Besides this, the industrial norms separate the two spaces occupied by a human and a robot by means of physical fence or wall [9]. Therefore, the success of such physical HR interaction is possible if the robot is enabled to handle this interaction in a smart way to prevent injuries and damages.

Wheeled mobile robots have attracted a lot of interest recently due to useful practical tasks such as robot fork trucks handling palettes in factories or obstacle avoidance. Based on the location of the load, a trajectory must be generated which ends precisely in front of, and aligned with the fork holes. The robot velocity must be zero at terminal point. It is often practically desirable that the mobile robot should avoid an obstacle while staying on the road. In such a case the robot should approach the obstacle as close as possible, which may lead to the necessity of determining the collision-free paths of minimum lengths. The aforementioned tasks belong to a class of motion planning problems. Motion planning is concerned with obtaining open loop controls which steer a platform from an initial configuration (state) to a final one, without violating the nonholonomic constraints. In addition, control constraints resulting from the physical abilities of the actuators driving the wheels should also be taken into account during the motion. Moreover, if there exist obstacles in the work space (state inequality constraints), the controls should steer the robot in such a way as to avoid collisions with obstacles. In such a context, two basic approaches to solving this problem may be distinguished. One is based on global methods of nonholonomic motion planning, which may be divided into discretized and continuous. The discretized technique is a sequential search of a graph whose edges are generated based on a discretized control space. Graph search methods proposed in [1, 2, 3] generate the globally optimal motion. The drawback of using graph-search techniques for trajectory generation is the resolution lost due to discretization of state and/or control space. Global continuous methods may be categorized into two classes. The first class uses an optimal control theory to generate robot trajectories. The Pontryagin maximum principle has been used in [4] to determine the time optimal trajectory in the unobstructed work space. The resulting controls are discontinuous and bang-bang. Using the calculus of variations and parameterization of controls, works [5, 6] convert the continuous optimal control formulation into an equivalent nonlinear programming problem. The second class of nonholonomic motion planning techniques offered in works [7, 8] is based on the use of Newton’s method. Its adaptation to collision avoidance is contained in [9]. However, it requires a time consuming iterative computational procedure. Moreover, only local optimization of a performance index is carried out when searching for the robot trajectory. A near realtime optimal control trajectory generator is presented in [10] which solves eleven first-order differential equations subject to the state constraints. Application of averaging method to motion planning is presented in [11]. The author of [11] proposed the use of truncated Fourier series to express the steering input and found the solution based on a shooting method. A suitable transformation of nonholonomic constraints and trajectory parameterization has been proposed in [12] to avoid obstacles. Nevertheless, this method does not take into account control constraints. In the context of mobile manipulators with nonholonomic platform, motion planning algorithms at the control feedback level have been proposed in works [13, 14]. The second approach to the nonholonomic motion planning is based on local methods. Among them, a method based on a Lie algebra of system evaluated at a given point is the most representative. It has been developed in work [15]. However, the computation of the GCBHD formula [16] seems to be time consuming. In order to adequately react to the environment, based on sensed information gathered during the robot movement, a nonholonomic motion plan must be generated in real time. Such amotion (desired trajectory) is simultaneously provided to online controllers (see e.g. [17, 18]), which may considerably increase precision control. It is obvious that almost all the aforementioned algorithms are not suitable to realtime implementations due to their computational complexity.

In the wide field of vehicle autonomous navigation, significant research efforts have been dedicated to the problem of optimal motion planning. The work presented in this paper faces the aspect of minimum-time velocity planning in the context of the so-called path-velocity decomposition [6] and the iterative steering navigation technique [8,5]. The robot vehicle has to travel on an assigned geometric path and the vehicle velocity on that path can be determined by assigning a minimum-time motion with arbitrary velocity/acceleration boundary conditions. Consequently, the relevant velocity planning problem is the synthesis of a velocity

C

1

-function that permits in minimum-time and with a bounded jerk to interpolate given velocity and acceleration at the time planning interval endpoints and to travel a given distance. The condition on the maximum jerk value permits to obtain a smooth velocity profile [3].

In many motion planing problems, the feasible subspace becomes thin in some directions. This is often due to kinematic closure constraints, which restrict the feasible configurations to a lower-dimensional manifold or variety. This may also be simply due to the way the obstacles are arranged. For example, is planning for a closed chain different from planning a sliding motion for a washer against a rod? An illustration for the two instances of such problems is shown on Fig. 27.1. Traditionally, the two problems are solved with different methods in motion planning. However, the two seemingly different problems have similar algebraic and geometric structure of the sets of feasible configurations. In either case, there may exist functions of the form |

f

i

(

q

)| ≤ 

ε

i

,

ε

i

 ≥ 0, that contain most or all of the feasible set. Typically,

ε

i

is very small, and in the case of closed chains,

ε

i

 = 0. When this occurs, a region of the feasible space has small

intrinsic

dimensionality, in comparison to the

ambient

configuration space. Note that in the problem on Fig. 27.1(b) and in similar motion planning problems, the

f

i

are not necessarily given.

Motion planning is one of the most studied problems in robotics. Various methods for solving this problem have been introduced in the last two decades. Applications beyond robotics including 3D object manipulation, computational biology, computational graphics, or drug folding are presented in [10].

During the last decade the RRT algorithm [11] has become widely used for solving the motion planning problem. The algorithm is based on random sampling of a configuration space. The sampled configurations are connected to a tree structure in which the result path can be found. The algorithm can be divided into three main parts: selection of a vertex for expansion, expansion and terminating condition. The original RRT algorithm is outlined in Algorithm 28.1.

The robot localization problem represents a key aspect in making a robot really autonomous. The position of the robot has to be estimated accurately based on the information about the surrounding world obtained from the sensors.

The current trend in this field [3] is to fuse relative and absolute measurements together. The aim of this is to provide a better position estimation of the robot location based on the differing nature of the data from different kinds of sensors. In order to fuse information from different sources, Kalman filtering is used, which is one of the most widely applied methods [2]. This approach formulates the localization problem as a state-estimation one.

Autonomous navigation in an outdoor environment can be effectively performed by small robots but remains a challenging task as soon as real scale ground vehicles are concerned. An intelligent mobile robot which uses a real armor ground vehicle as a platform is currently being developed for security applications (Fig. 30.1(a)). The robot tasks needed to be addressed are patrolling, surveillance, and tracking suspicious objects in a limited area. Executing those tasks a security robot often faces the situation when many objects in the environment have similar geometrical characteristics. This situation is typical for navigating within the outdoor warehouse [1] as illustrated in Fig. 30.1(b). The cargo containers (the squares with pattern) of the same size are piled inside the fenced area. A robot is used to carry out the patrol along a given trajectory (dotted line) around the area. Being equipped with the basic positioning system using GPS receiver and odometry sensor, a robot is able to follow the given path if measurement no errors are negligible. In practice, due to various errors (misalignments, drifts, slippage) basic GPS/odometry-based system fails to provide reliable pose estimates required in physically restricted environments. Measurements of relative distances are necessary and they are provided in our case by an additionally installed laser range sensor (LRS) known for its accuracy and reliability.

Autonomous systems have received special attention in the scientific community in the hope of designing a mobile robot with the ability to navigate in an unknown environment and simultaneous localization and mapping (SLAM) problem has been considered as a fundamental issue to realize the above task. Consequently, the robotics community has been engaged in the development and implementation of algorithms and related issues for SLAM. A lot of effort was put into solving this challenging problem [1, 2, 3, 4]. The widely used formulation of the SLAM problem has been the augmented state approach in an estimation theoretic framework, generally based on the extended Kalman filter (EKF) and using the relative observation only to update the state vector [5, 6, 7].

A camera attached to a mobile platform allows the extraction of information about the environment and the robot itself. The acquired video can be processed to estimate the translational and rotational velocities on each axis of the camera, which are directly related to the velocities of the platform, producing a sensor that relies on the vision system, useful in situations where traditional sensors are no longer reliable.

The movement of the camera produces a variation of the brightness pattern of the image, so the optical flow can be estimated. It can be shown that this flow is the result of the sum of the patterns associated with the movement in each degree of freedom. When each pattern is known, the movement components associated with each DOF can be calculated.

A number of robotic publications, some of which utilize information available in acoustic signals, deal with the navigation of robots. In 1986, J. Schoenwald

et al.

developed a technique that utilize acoustic sensors to adapt the robot path in a changing environment [8]. G. Tesch’s work is based on analyzing echo signals produced in an unknown, unprepared environment by sound sources on the robot itself [9]. In a most recent scheme presented by [10], a robot generates acoustic signals and a microphone array determines its location by capturing these signals.

Over the last 50 years one has been able to observe enormous enhancement in positioning and measuring accuracy. The main portion of this enhancement is the result of improved knowledge about high precision machine design. A fundamental principle was recognized by Professor Abbe already in the 1890s about the aligment of the displacement measuring system with the distance to be measured. Another fundamental principle is the separation of the structural and measuring functions in a machine. As mechanical accuracy is costly, unlike repeatability, software techniques were used from the beginning to compensate for the systematic errors in order to keep manufacturing costs low.

The considerable progress made in manufacturing has made it possible to construct industrial manipulators with more and more stiff joints. This presents an obvious advantage in an unlimited space, which results in increased positioning accuracy. However, problems arise when the robot is equipped with a torque and force sensor for the purpose of a controlled action on the environment, for example in the tasks of autonomous assembly. In such cases the high environment stiffness may account for generation of big forces which have a detrimental effect on the mechanical construction of the manipulator and constrain the allowable robot velocities in cases when the robot comes into contact with the environment [3]. Therefore, passive [18] or active [16] compliance systems are introduced between the stiff manipulator and also the mostly stiff environment. They are usually added as a part at the end of the robot arm. A passive compliance uses some mechanical devices composed of springs, sliding axles and knee joints. In passive compliance devices, there is no interaction between the control computer of the robot and the compliant device. The chief drawback of such a solution is lack of information about the exact position of the object in the space [1]. Since precise positioning is a matter of prime importance in the majority of applications, active compliance systems equipped with sensors coupled with the robot control system take on great significance. One such commercial example is the DLR (German Aerospace Center) lightweight manipulator arm with torque sensors in each joint along with collision detection and safe reaction system [2]. Although the active compliance systems based on the hybrid position-force control are much more attractive than the passive solutions based on the impedance control, they present a great challenge due to occurring stability problems [20]. Additionally, it should be remembered that the presence of a passive compliance between the manipulator and the environment is required for entering into contact to be noninvasive, even if an active compliance is used [19].

Many climbing robots are specialized for a certain substrate like glass-like Stickybot [20] (suction mechanisms or adhesive mechanisms) or porous substrates like Spinybot [21] (claw-like mechanisms). Thus, these robots are often expensive mechatronical systems. One possibility to gain more flexibility in application and to reduce costs is to modularize such systems. On the one hand we have to achieve high flexibility in application, but on the other hand we need a high performance robotic system. A major challenge is to do the step from specialists to generalists in climbing robots.

Common goals for mobile robots are locomotion, ideomotion, manipulation, navigation, orientation, imitation, and cooperation. For a climbing robot, the demand in locomotion is high locomotion ability. Climbing is locomotion by generation of directed body motion and of substrate contact in alternating combination. The robot has to move horizontally and vertically, ascending and descending – an interesting challenge for biologically inspired robotics.

Thanks to highly sophisticated control techniques, modern mobile robots are capable of accomplishing advanced tasks. Briefly, mobile robots can be dived into two main branches: research devices and consumer devices. The firs class usually consists of prototypes used for testing and verification of various control algorithms. The latter is represented by efficient, dependable and fulfilling industrial safety requirements robotic vehicles, often distributed as closed systems. During the design and test phase, the imperfections of a completely custom design may not allow to perform advanced tests, thus lengthening the construction phase. In order to eliminate this problem, the authors have designed the robot so that its crucial parts were previously known, tested and proved to be reliable. The control system consists of distributed pieces of software and hardware, each of which accomplishes precisely defined tasks. A fast serial communication bus is used to exchange information, what makes possible the use of module elements of control system. The modularity of the control system in range of hardware and software allows easy modification of components of the control system (the exchange), the diagnostics of drives as well as the introduction of an alternative controller or a measuring sensor in case of breakdown.Moreover the presence of a communication bus makes it possible to integrate the control system with measuring devices and drives. The utilization of independent arrangements of control drives makes it possible to use synchronous drives with position or speed control, independently from the main driver, which enlarges control precision [3]. The robot served as a testbench for various control algorithms thanks to which it was possible to determine the effectiveness of the design.

Recently there have been many large-area objects that require supervising in a systematic manner. The supervision is needed due to complexity of the systems included in the objects, such as power, communications, production, transport systems, and many others. There are many reasons why the objects deserve supervising. On the one hand, the objects themselves and the equipment remaining in the area can be very expensive, or be property of a third party. On the other hand, the systems can play essential role in providing populations with water, energy, oil, gas, heat and other resources. Finally, the media stored in the area or transported within the systems can be dangerous to the environment or even can be applied in improper way, the last being especially important in the era of terrorism. All these arguments support a vital need of constant supervising such objects. This task is usually done by specialized, well-trained and equipped humans.

It is forecasted that by the year 2013, a quarter of the people in Japan will be over the age of 65. An increase of the elderly population having physical function problems due to diseases and advanced age has been one of the most serious problems in Japan. In the sites of occupational therapy facilities, simple training tools, such as sanding boards, ropes, ring toss and peg boards are widely used. For this reason the therapeutic effect depends on the therapist. Moreover, there is no evidence of the treatment results, because the subject’s dynamic data is not acquired and the therapeutic effect of the training is not evaluated quantitatively. So, the occupational therapy is called empiric treatment. Moreover, the working efficiency of a training by the therapist has to increase, because the treatment time is limited.

Since the beginning of the 1990s robots have been successfully implemented in medicine. The most advanced are telesurgical systems. In this group only two constructions were used in practice, i.e.:

Zeus

® Robotic Surgical System produced by Computer Motion [1] and

da

Vinci

® Surgical System by Intuitive Surgical® [5, 2]. However, after the patent litigation in 2003 Intuitive Surgical® and Computer Motion merged into one company and now only one –

da

Vinci

® Surgical System – is produced.

Its large cost, which is about 1.000.000 EUR, makes it unavailable to Polish hospitals for financial reasons. This was why the attempts to prepare similar constructions were undertaken in Poland. However, it was decided not only to copy but also to undertake the challenge of constructing an improved version of the robots mentioned above, at least in some aspects.

Walking ability, though important for a quality of life and participation in social and economic life, can be adversely affected by neurological disorders such as spinal cord injury, stroke or traumatic brain injury. The main goal of rehabilitation is to maximize motor performance and minimize functional deficits [4]. This requires a patient to practice repetitive motion, specifically using the muscles affected by neurological injury. In traditional rehabilitation, one or more therapists assist and encourage the patient through a number of repetitive exercises. Nowadays in a rehabilitation, mechanical systems (robots) are used to replace or assist the lost functions of human extremities and an attempt will be made to present them in more precise terms than hitherto.

How can we build systems that have the kind of useful autonomy that would be acceptable to real-world, non-technical users? That is a fundamental question facing the special breed of engineers who are not just visionary but also sufficiently reckless to attempt driving innovations borne in research and development worlds to the market. In this paper we share our first hand experiences and observations from the ultimate adventure of bringing an autonomous delivery robot to real life.We believe that some of lessons we have learned are sufficiently general to be applicable to a range of relevant cases.

The interactive behavior operated trolley (InBOT, Fig. 43.1) is an approach to transfer state of the art robotics technology into human everyday environments. The benefits of the actual technology shall become available for the masses. InBOT addresses several everyday problems and therefore the main task is to ease the daily shopping. Among other possibilities this means to help the customer to find the desired products without extensive search in big supermarkets or to relieve the customer from the burden to push the shopping cart using his own force all the time, especially if the cart is heavily loaded or the customer is elderly or handicapped. InBOT provides four different modes of operation. These are

haptic steering, following, guiding

and

autonomous

. In all the modes the user is assisted in several ways. Additionally InBOT is able to perform local maneuvers like finding a parking position at a wall or turning to help loading heavy goods. A flexible mode-dependent task-planner rounds out the capabilities of InBOT. It is able to stop and continue or postpone given tasks at runtime in combination with topological navigation [8]. A rough overview over the main components of InBOT is illustrated in Fig. 43.2. Finally, InBOT is equipped with the holonomic mecanum drive because the robot shall not lack the manoeuvrability of conventional shopping carts.