DESIGN OF A HOLONIC SELF-RECONFIGURABLE ROBOTIC SYSTEM A. E. TURGUT1 M. DURNA2 İ. ERKMEN3 A. M. ERKMEN4 A. ERDEN5 Middle East Technical University 1. Mechanical Engineering Department. E-mail:
[email protected] 2. Electrical-Electronics Engineering Department. E-mail:
[email protected] 3. Electrical-Electronics Engineering Department. E-mail:
[email protected] 4. Electrical-Electronics Engineering Department. E-mail:
[email protected] 5. Mechanical Engineering Department. E-mail:
[email protected] 06531-ANKARA TURKEY key words: self-reconfiguration, holonic, modular robotics, distributed control
ABSTRACT This paper investigates the holonic modeling of a self-reconfigurable multiple module robotic system which operates in a two dimensional space. Holons are the simplest physical elements of the system, which have the capabilities of independent actuation, and of cooperation with other holons in achieving global behavior for the whole colony structure. Our work focuses on the concept of self-reconfigurability of the modular colony structure so as the holons reconfigure themselves to a pre-described structure in the absence of a human in the loop. The reconfiguration control of the dynamical holonic structure is developed based on hybrid control, since our system exhibits both discrete and continuous behaviors. This paper introduces the mechanical design of holons, their connection/disconnection and motion mechanisms together with the control architecture developed. Acknowledgement: The holons considered in this work are being manufactured by supports provided by project no: AFP-97-03-02-01 We would like to extend our appreciations to Mr. Ahmet H. Alkan and to ODESA Electromechanics Ltd. for their valuable helps on the implementation of the robots.
1. INTRODUCTION Holons, as first introduced by Arthur Koestler[1], usually work in cooperation with other holons of a group, forming a modular flexible system. The basic property of a system having a holonic architecture is of being built from simpler components working in colony so as to achieve a global behavior of a higher order [1,2]. In holonic robotic structures, the overall system behavior is determined not only by the individual autonomy of each primitive robot subsystem that builds up the whole structure, but also, by the
cooperation between those subsystems. Distributed decision-making and control handle cooperation in such systems. Many works exist in the literature using modular systems that pertain to the principle of independence and cooperation, although not all called holonic. For completeness we provide here a brief review of such approaches. I-Ming Chen [3] studied the optimum configuration of assemblies built up from modular robots for a given task. He considers kinematic properties of links in cubic or prismatic geometry joined by joint elements, whose optimum taskoriented kinematics is obtained using genetic algorithms due to the discrete nature of the problem. Fukuda et al [4] focused on robotic systems they call CEBOTs (Cellular Robotic Systems) which are functionally different robots, working together in the completion of a given task by sharing subtasks. Fukuda et al make use of Genetic Algorithms in the trajectory planning of each robotic module called a cell. Gregory Chirikjian [5] designed his own modular robotic system and produced few prototypes. The main points in his work are the strong couplings between the modules producing stiff structures that can also accomplish reconfiguration. Because of the self-reconfiguration ability of these systems, he calls them “metamorphic”. These modules are designed in a hexagonal geometry due to the fact that symmetry brings a great simplification to the analysis and control of these self-reconfigurable systems. Chirikjian uses the simulated annealing method to decide on the action of each module in the reconfiguration phase. Similar to Chirikjian, S. Murata works on the self-reconfiguration of modular structures made up of symmetric modules. Since their hexagonal symmetric structure is similar to fractal structures, he calls each of the modules as a “fractum” [6]. As it is also the case for the metamorphic robots
of Chirikjian, fracta move within the structure by “sliding” over each other. In both of the works [5,6], the coupling principles of the modules are the same and are based on electromagnetic fields. The need of holonic systems arises from the unstructuredness of the media where the robot has to operate and from the changing nature of the task in that medium. In order to adapt to such a medium the system must change its structure in an optimum way so that a variable structure is achieved in the control and task spaces while keeping itself robust and accurate in behavior. In this paper, we aim at introducing the design and control of a flexible modular robot system that reconfigures its dimensionality and kinematics according to a given task, as opposed to conventional ones which have fixed physical and kinematic properties after being manufactured.
2. REQUIREMENTS AND ASSUMPTIONS FOR CONTROL ARCHITECTURE Our work focuses on the autonomous, continuous control of holonic reconfiguration called selfreconfigurability. Reconfigurability which is “the ability of a robot to rearrange its own physical components” [7,8], is extended in our work to selfreconfigurability that enables the modular robot structure to reconfigure itself for a given task without any external help. Autonomous control of holonic reconfiguration is achieved in two hierarchical levels: Holonic Control and Colony Control.
2.1 HOLONIC CONTROL In our work holons are physically and functionally identical, so that they are “unit modular”. Unit Modularity is described as a system, which is composed of a single type of repeated components. Each holon in this architecture is able to communicate with each other and a host system. They also have the ability of connecting and disconnecting with the neighbouring holons subject to kinematic constraints implied by the design. The ability of these holons to move over each other in the locomotion phase changes the overall colony configuration.
2.2 COLONY CONTROL A colony is able to modify its kinematic properties by changing connectivity between holons involved in the colony. While holonic control is continuous in nature, we model colony control as a discrete control on Colony Configurations. This discrete control is formalized as: From an initial Ci to a given predefined final configuration Cf , find a minimal length sequence of discrete transitions called a ‘run’, σ = σ1•σ2• . . .
σ
such that Ci → Cf where • is the relational composition operator and each σi ∈Σ, i∈N+. Here Σ is the set of discrete transition events. This Configuration Flow Graph (CFG) is subject to the constraints on the transition events. Run length is finite due to finiteness of the generated configuration space. We represent the connectivity of holons in a colony as an incidence matrix. The colony configuration represented by an incidence matrix is used to search a configuration space and generate the “Kinematical Flow Graph” (KFG). This name is due to the kinematically different configuration nodes of that graph we generate where each adjacent node can be obtained by an addition or removal of a vertex. Clearly, this graph cannot be directly used for a flow in configuration space because of the loss of knowledge on weights (i.e. norms of families of the configuration) of the edges. However if two nodes C1 and C2 are adjacent in CFG σ
(i.e. there is a transition event σ∈Σ, such that C1 → C2) then the modified configurations obtained from those two original configurations is adjacent in KFG. The relation between those adjacent nodes in KFG is the addition of a vertex if originally the transition is “connection”. For an illustrative example of colony control applied to holonic grasping in our earlier works, the reader is kindly referred to [9]. The present paper, on the other hand, is based on the design of holons and their holonic control with an emphasis on the mechanisms of connection and disconnection as well as on the communication architecture developed. Fig. 1 introduces the hierarchical control and communication used in our work. We allow human intervention only at the highest level where system behaviour and control can be monitored. The host converts the user defined control tasks into holonic sub-control tasks that are input to the continuous control at the lowest level of the hierarchy and which possesses a distributed nature. The host embodies the discrete colony control architecture and generates the optimal path in a Kinematical Flow Graph by finding an optimum sequence of colony configurations.
3. DESIGN OF A HOLON Each holon is a mechatronic device that has the following properties: a. It is equipped with the minimum number of actuators and sensors found necessary for its locomotion about its neighbours. The holon also embodies mechanisms that enables it to connect to and disconnect from other holons. b. It possesses a set of processors that constitutes the intelligence needed for holonic control and communication.
The active port is the male one, which is composed of a push-pull type solenoid that activates the connector block. The female port is just a guideway in which the connector block of the male port fits. In the rest position, the connector block is in its farther most location. When a connection is to be made the connector block is pulled by the solenoid and it is then released when female port of the opposing holon is aligned. As the male port is passive in the rest position very little amount of power is consumed in the overall movement of a holon regarding this sub-system.
3.3 COMPUTATION AND COMMUNICATION ARCHITECTURE
FIGURE 1 HIERARCHICAL CONTROL STRUCTURE In this section, a detailed explanation of the two functionality of our holon, namely its motion and communication issues will be given.
3.1 ACTUATION SUB-SYSTEM The actuation sub-system consists of two dcmotors, one shaft encoder, one rack-pinion mounted on a rotating table as seen in figure 2. Motor 1
Middle Plate
Motor 2
Rack and Pinion
Rotating Table
FIGURE 2. ELEMENTS OF LOCOMOTION SUBSYSTEM The table is rotated by motor 2, which is perpendicular to the plane of the figure, and the pinion on motor 1 moves the rack back and forth. Both of the motors are controlled by an on/off strategy. The difference among them is that feedback for motor 1 is taken from an encoder coupled to it, whereas the feedback for motor 2 is taken from a roller type micro-switch mounted on the rotating table and the micro-switch slides on the middle plate. Middle plate has eight 45o apart holes in which the roller of the micro-switch can fit into. In this way, the table can be rotated at steps of 45o, which eliminates the use of complicated PID algorithms.
3.2 CONNECTION SUB-SYSTEM The connection sub-system generates the physical connection between holons. Its mechanism is composed of two male (+) and two female (-) connection ports as seen in figure 4.
At the top level of the whole system a host PC broadcasts the necessary low-level holonic control commands for the holonic structure via a wireless communication channel. These commands are generated during colony control when the next configuration is decided for the colony through the sequence in the KFG. Infra red transmitters and receivers are used in that part. The frame of a message in the whole system is given in Fig. 3. The format of the command line and the whole set of commands sent by the host via IR channel to the colony is given in Table 1. Although different communication schemes (IR, I2C, and SPI) are used in the system, the message frame is unique. Local Map, seen in Table 1,is a 3x3 matrix that keeps the knowledge of neighbouring holons with the holon ID’s given in the entries. ID=0 is reserved for host and ID=255 is for NULL neighbours. The entries of the matrix are defined via a direction convention that can be seen in Figure 7. As the flexible modular robotic structure is a dynamic network of interconnected holons and the connectivity scheme is changing over the time, the design of the communication architecture is an important issue. We have stated earlier that each holon should have at least one channel of communication with one of its neighboring holons. For the supervisory control of the Host, it is a must that a unique identification be assigned to each holon of a given colony configuration. This means that broadcasting a message thorough a priority based law will be adequate for the communication that exists between any two holons in a colony configuration.
FIGURE 3. FRAME OF A MESSAGE IN THE NETWORK The frame format used in the Host-Holon communication is given in Fig. 3. The same frames are also used in the intra-colony communication. For the
message transmission. “Slave” holon follows the commands issued by the “Master”. The discrete command (which results in a discrete transition discussed above in Section 2.2) is DC which means “DisConnect”. The other commands (i.e. Counterclockwise Rotation-CCW and Clockwise Rotation-CW ) are continuous in this sense and they reside between two DC commands and do not trigger a colony configuration change. This structure is the one which forms the backbone of the hybrid control system theoretical framework.
intra-colony communication, the holons of the same colony that is being controlled by the host at the top will talk with each other. The intra-colony communication messages have a format where the sender and receiver identification (ID) code exists together with the control command as a header to the data. But, for this time naturally the command set used is different and is given in Table 2. In the following discussion, the term “Master” indicates the holon which is about to take continuously controlled action toward a discrete transition of the colony configuration. Master is the one, which initiates the Male Connector Block in rest position
Male Connector Block in pulled active position
Solenoid
Female Connection Ports
FIGURE 4. ELEMENTS OF CONNECTION SUB-SYSTEM Receiver Holoni
Description Initiates the Message transmission Terminates the Message transmission Go To specified location given by dir. Local Map of receiver with neighbouring I.D.s defined by Mi’s. Host Holoni Task Accomplished. Result of the last G0 TA command. =1 on success, =0 on failure. TABLE 1. INSTRUCTIONS AND COMMAND FORMAT USED IN HOST-HOLON COMMUNICATION
Receiver Slave Holon
Sender Host
Sender Master Holon
Command IM TM G0 LM
Command DC CCW CW TA
Description Slave pulls connector block located in direction Slave rotates its table to direction in CCW. Slave rotates its table to direction in CW. Result of the last CCW or CW command. Master Slave =1 on success, Holon Holon =0 on failure. TABLE 2. INSTRUCTIONS AND COMMAND FORMAT USED IN HOLON-HOLON COMMUNICATION The communication type of the holons is selected as 2-wired bus communication with the transmission protocol I2CTM -Bus by Philips Corp. So each data byte of the message is transmitted and received serially via the communication channel in the I2C format. The rate of data transfer on an I2C bus can
reach up to 400kbit/s. Data and clock buses are configured as given in Fig. 5. The processing unit of each holon has bus terminations at each connection port such that they remain connected to the relevant terminals of the neighbor from that side. Such a mechanism solves problems that arise from the
dynamical nature of the whole system since this mechanism, groups all the bus wires into a node where holons are directly connected. The processing unit of a holon is built up from two processors connected to each other with an other serial bus named SPI (Serial Peripheral Interface) which can reach up to 1.5 Mbit/s. The main processor which incorporates the operating system equipped with task manager functions, is a MC68HC11 embedded microcontroller. The second processor is a PIC16C74, which is responsible of low-level functions. The tasks which HC11 is responsible of, are to drive motors, receive messages from the IR channel, decode them into low-level subtasks, prepare the messages to be sent to neighbouring holons and receive broadcasted messages via PIC. PIC controller acts as a slave of HC11 for driving solenoids and being an interface for the communication between the main processor of that holon and that of the neighbouring holons.
3 2
1
4
5
FIGURE 6. HOLONIC STRUCTURE d1 d0
d2
d7
d3
d4
d6 d5
FIGURE 7. THE DIRECTION CONVENTION OF A HOLON
1
1
1
FIGURE 5. COMMUNICATION SCHEME OF HOLONS.
2
2
a
4. LOCOMOTION: A TYPICAL EXAMPLE A typical change of the colony configuration is given in figure 6 as monitored by the Host metacontroller. There, shaded modules 1, 3 are candidates to move about their neighbouring holons numbered as 2, 4, 5. Holon 1 couples with holon 2 via its rack and pulls its male connector block so that it releases the connection with holon 2 (Fig. 8a), then it is rotated by holon 2 90o clockwise (Fig.8b, 8c, 8d). Finally holon 1 rotates itself 90o counter clockwise (Fig. 8e, 8f) so that it aligns its female port with male port of holon 2 then a new connection is formed between the male and female ports.
c
b
1
1
1
d 2
2
e 2
f 2
FIGURE 8. MOVEMENT OF HOLON 1
5. CASE STUDY: A 2-HOLON COLONY In this section, a detailed explanation for the whole process including the intra-colony (holon-to-holon) and host-to-holon communication together with the locomotion issues will be given for a colony composed of two holons. The colony is given in Fig. 8. The colony in Fig. 8a constitutes the initial configuration Ci. The final configuration Cf is given in Fig. 8f.The problem is to reach Cf from Ci. The Host-Holon and intra-colony communication during a transition is given here as an example. The holon to be actuated to a specified location is Holon 1. The following example is formatted based on Tables 1 and 2. Therefore tracking the example should be done by referring to those tables.
Steps
Receiver Holon ID
Sender Holon ID
Command
Description
1 2 3
1 1 1
0 0 0
Steps
Receiver Holon ID
Sender Holon ID
5 6
Holon 1 turns its rotating table to direction d3. 2 1 CCW d7 Holon 1 acts as a master, requesting holon 2 to turn its rotating table counter clockwise to direction d7 1 2 TA 1 Requested action is completed Holon 1 moves its rack forward to fit into Holon 2’s rack slider Holon 1pulls its male connector block so that releases the connection with Holon2. 2 1 DC d1 Holon 1 requests Holon 2 to pull its male connector block to prevent male block to block its way while rotating 2 1 CW d1 Holon 2 is requested to turn its rotating table to position d1 in the clockwise direction. 1 2 TA 1 Requested action is completed. Holon 1 rotates itself 900 in CCW direction to align its female port with male port of holon 2 Holon 2 releases its connector block back to rest position. Holon 1 releases its connector block to rest position. Holon 1 moves its rack back. 0 1 TA 1 Acknowledge the Host.
IM Host (ID=0) initiates transmission with holon 1 G0 d2 The receiver holon (ID=1) is requested to go dir= d2. LM [255, 255, Local Map (defined in Section 3.3) is downloaded to 255, 2, 255, 255, ;receiver 255, 255] 4 1 0 TM Transmission channel is closed by host. By this sequence the command and local map is transmitted to the holon 1. From that point on, the intra-colony communications and actuations take place.
7 8 9 10
11 12 13 14 15 16 17
Command
Description
TABLE 3. PROCESS STEPS OF MOVEMENT OF A HOLON IN A 2 HOLON COLONY
6. CONCLUSION In this paper a holonic robotic system is introduced in terms of its hardware design as well as its communication protocol and control architecture. The rack and solenoid mechanisms provide more flexibility and accuracy in locomotion when compared to existing holonic systems. In the colony structure, communication traffic is dense while local holonic control is a simple feedback control. The complex problem of colony configuration control is handled with the powerful processing centre of the Host by discrete control. The communication issue is simplified using a top-down abstracted protocol.
7. References [1] M. Hirose, Development of the Holonic Manipulator and Its Control, Proceedings of the 29th Conference on Decision and Control, Honolulu, Hawaii, December 1990. [2] Ling Gou, Tetsuo Segawa, Peter B. Luh, Shinsuke Tamura, John M. Oblak, Holonic Planning and Scheduling for a Robotic Assembly Testbed, Proceedings of the 1994 IEEE Int. Conf. on Robotics and Automation, pp. 142-149. [3] I.-M. Chen and J. W. Burdick, Enumerating Non-Isomorphic Assembly Configurations of
View publication stats
Modular Robotic Systems, 1993 Proc. of IROS, Yokohama, Japan, pp. 1985-1992. [4] T. Ueyama, T. Fukuda, F. Arai, Coordinate Planning using Genetic Algorithm - Structure Configuration of Cellular Robotic System-, Proceedings of the 1992 Int. Symposium on Intelligent Control, Glaskow, U.K. [5] G.S. Chirikjian, Kinematics of a Metamorphic Robotic System, Proceedings of the 1994 IEEE Int. Conf. on Robotics and Automation, San Diego, CA, May 1994, pp. 449-455. [6] S. Murata, et al, Self-Assembling Machine, Proc. Proceedings of the 1994 IEEE Int. Conf. on Robotics and Automation, San Diego, pp. 441-448. [7] Mark Yim, New Locomotion Gaits, Proc.of IEEE Int. Conf. Robotics and Automation, pp.2508. [8] Mark Yim, Locomotion with a Unit-Modular Reconfigurable Robot, PhD Dissertation, 1994. [9] M. Durna, I. Erkmen and A. M. Erkmen, Holonic grasping, Proc.of the IEEE/RSJ Int. Conf. On Intelligent Robots and Systems, Canada, October 1998, Vol. 1, pp.140.
