PROVIDING SPATIAL INTEGRITY FOR DISTRIBUTED

UNMANNED SYSTEMS

Peter Simon Sapaty

Institute of Mathematical Machines and Systems, National Academy of Sciences, Glushkova Ave 42, 03187 Kiev, Ukraine

Keywords: Unmanned systems, Distributed scenario language, Networked interpretation, System integrity, robotic

swarms, Reconnaissance, Camp security, Convoys, Explosive ordnance disposal, Gestalt, World super-

machine.

Abstract: Due to the increased complexity of tasks delegated to unmanned systems, their collective use is becoming of

paramount importance for performing any reasonable jobs. An approach is offered where group behaviors

are accomplished automatically rather than set up manually, as usual. Missions in the Distributed Scenario

Language (DSL) can be executed jointly by communicating interpreters in system units. Scenarios like

reconnaissance, camp security, convoy, mule, and explosive ordnance disposal in DSL, oriented on different

numbers of cooperating units, are demonstrated. The approach allows us to effectively manage any teams,

from human to robotic, and from homogeneous to heterogeneous, regardless of the number of components

in them. A variety of other applications of the technology are outlined too, already researched or prospective,

also its relation to the gestalt philosophy, where super-summative whole dominates over system parts,

defining their sense and even existence, rather than vice versa. The paradigm discussed may also represent a

distributed dynamic world super-machine operating in parallel with both information and physical matter.

1 INTRODUCTION

With the world dynamics increasing due to global

warming, numerous natural and manmade disasters,

military conflicts, and international terrorism, using

unmanned (ground, sea, underwater, and air)

systems can alleviate many problems and save lives

in hazardous environments. Because of the

complexity of tasks delegated to unmanned solutions

and still insufficient capabilities of existing robotic

vehicles, their simultaneous, collective use may be

of paramount importance to perform any reasonable

jobs. Operating together, the unmanned groups,

often called swarms, can fulfill the required

objectives despite possible runtime damages to

individual units.

We are offering a novel approach to organization

of unmanned systems, oriented from the very

beginning on parallel solutions in physical spaces,

with swarm behaviors resulting naturally and

accomplished automatically, rather than

programmed manually. This approach, symbolically

called “overoperability” from the previous

publications (Sapaty, 2002, 2005), allows us to

create, modify, analyze, process, and manage any

distributed systems, establishing local and global

dominance over them.

Within the overoperability philosophy, an

integral mission scenario, written in a special high-

level scenario language (Sapaty, 1999, 2005) and

reflecting semantics of what to be done in a

distributed space rather than details of

implementation, is executed in a parallel and

cooperative manner by dynamically networked

unmanned units. During the scenario evolution, any

operations can be accomplished in the world, along

with the needed movement of code, equipment and

“doers” (both artificial and biological), as well as

creation and maintenance of physical and virtual

infrastructures supporting the missions.

This paper is essentially inspired by the

European Land Robotic Trial (M-ELROB, 2008) in

which the author participated. It was conducted to

provide trials as close as possible to operational

scenarios for UGVs/UAVs with focus on short-term

realizable robot systems. The day and night trials

were organized within the following five main

scenarios: non-urban reconnaissance, camp security,

transport convoy, transport mule, and explosive

ordnance disposal. Only a limited number of robotic

units were engaged in every scenario, just one or

Sapaty P. (2009).

PROVIDING SPATIAL INTEGRITY FOR DISTRIBUTED UNMANNED SYSTEMS.

In Proceedings of the 6th International Conference on Informatics in Control, Automation and Robotics - Intelligent Control Systems and Optimization,

pages 29-40

DOI: 10.5220/0002181200290040

 SciTePress

two, whereas every scenario could potentially be

executed with much higher efficiency if using

robotic teams with many units, which cooperate with

each other.

The paper also reflects activity on the project

started under the sponsorship of Alexander von

Humboldt Foundation (AvH) in Germany. One of its

aims is formalization of known mission scenarios in

such a way that they could be performed by any

available numbers of robotic vehicles, with the

management burden effectively shifted to self-

organized robotic teams -- thus relieving human

operators from traditional routines and allowing

them concentrate on mission goals and global

efficiency instead.

2 DISTRIBUTED SCENARIO

LANGUAGE (DSL)

The approach described here is based on the

Distributed Scenario Language (DSL), which allows

us to set what to do in a distributed world on a

semantic level, abstracting from details of how to do

this and with which resources, delegating these to

the intelligent automatic interpretation. Being a

universal programming language with advanced

parallel and distributed capabilities, DSL can also

describe tasks and behaviors on any levels, if

needed. The language can be used by humans who

should follow its instructions individually or

collectively, or can be directly executed by robots

and their teams. Any mixed human-robotic

organizations can be managed in DSL too.

2.1 The World

The world DSL operates with can be virtual,

physical, or combined.

 Virtual World (VW) is discrete and consists of

nodes and links connecting these nodes. Any

information can be associated with both nodes

and links in the form of their names (contents).

Nodes have unique addresses in VW, whereas

their names (same as names of links) may repeat

throughout the VW. Nodes can be accessed

directly, globally, by their names or addresses, or

locally, from each other, via the (named) links,

whereas links can be accessed only locally--from

the adjacent nodes. A variety of broadcasting

possibilities are available in the VW, both in a

global and local way, for example, from outside

to all nodes, from a node directly to all other

nodes, or from a node to all neighboring nodes

via the selected or all adjacent links.

 Physical World (PW) is continuous. Any point in

it can be identified and accessed by the

coordinates expressed in a certain coordinate

system, also with certain precision. Staying in a

PW point, you can lift local physical parameters

from the world and, possibly, also change them,

impacting the world locally too.

 Virtual-Physical World (VPW) is the one where

VW nodes additionally associate with certain

coordinates in the PW. VPW is discrete on a

snapshot, but the nodes can change their physical

coordinates overtime. The VPW nodes can be

globally accessed by their names, addresses, or

physical coordinates (for the latter, more

correctly: by coordinates of the expected center

and a radius of the region, due to limited

precision of the coordinates). Also locally, via

links--same as for the pure VW. In addition to

the broadcasting capabilities of VW, nodes in

VPW can also be massively accessed/entered by

identifying a probable region in PW where they

are expected to exist--by the region’s center and

a range (radius) from this center, where the latter

may be of any value.

2.2 High-Level Scenarios

The world, as described above, is navigated and

processed in a parallel and distributed way by high-

level DSL scenarios having the following main

features.

General Features:

 A DSL scenario (or program, in a conventional

notation) describes development of activities in a

distributed world as parallel transitions between

sets of progress points, or props.

 Starting from a prop, a program action may

result in one or more new props, or remain in the

same prop.

 Each prop has a resultant value and a resultant

state.

 Different actions (whatever complex they might

be), starting from the same prop, may evolve

independently or interdependently, and

sequentially or in parallel, each contributing to

the resultant set of props on this group of actions.

 Actions may also succeed each other in the space

of props, with new actions applied in parallel

from all props reached by the previous actions,

resulting altogether in the integrated set of props

on all these applications.

 Elementary operations can be defined on the

ICINCO 2009 - 6th International Conference on Informatics in Control, Automation and Robotics

values of props reached by other actions (the

latter of any complexity), leading to the resultant

prop with associated value (which may be

multiple) and resultant state.

 The scenarios can form new or remove existing

nodes and links in the distributed VW or VPW,

allowing us to create, modify, and process any

graph-based infrastructures in these worlds.

Association with World Nodes/Points:

 Any prop can be associated with a node in VW

or a position in PW, or both, like in the case of

VPW.

 A prop can also be linked separately with VW

nodes and PW positions, allowing us to operate

with the two worlds independently.

 Any number of props can be associated

simultaneously with same points of the worlds.

 Staying with nodes/positions in the worlds, a

prop allows us to directly access local data in

these points, both virtual (information) and

physical (matter).

Different Types of Variables:

 Heritable variables – these are starting in a prop

and serving all subsequent props, which can

share them in both read & write operations.

 Frontal Variables – are an individual and

exclusive prop’s property (not shared with other

props), being transferred between the

consecutive props, and replicated if from a single

prop a number of props emerge.

 Environmental Variables – are accessing

different elements of physical and virtual words

when navigating them, also a variety of

parameters of the internal world of DSL

interpreter.

 Nodal Variables – allow us to attach an

individual temporary property to VW and VPW

nodes; they can be accessed and shared by any

props associated with these nodes.

 Different types of variables, especially when

used together, allow us to create efficient spatial

algorithms which work in between components

of distributed systems rather than in them.

Hierarchical Control:

 DSL scenarios can use a variety of spatial control

rules, allowing us to assess local and remote

states, make local and global decisions, and

invoke or skip subsequent and terminate current

operations, on the results of these decisions.

 Nested control infrastructures, embracing the

whole scenario, provide interdependent local and

global decisions associated with proper points of

the worlds.

2.3 The Language Syntax

DSL has a recursive syntax shown below together

with names of its main constructs (where square

brackets are for an optional construct, braces mean

construct repetition with a delimiter at the right, and

vertical bar separates alternatives).

wave Æ constant | variable | [ rule ] ({ wave , })

constant Æ number | string | special

variable Æ identifier | reserved

rule Æ expand | transfer | modify | branch |

advance | repeat | grant | echo |

arithmetic | structural | assign | compare |

timing | type | usage | identifier | wave

out | in | infinite | nil | empty | first | last |

andom | virtual | physical | combined |

neighbors | global | local | direct

reserved Æ N{alphameric} | H{alphameric} |

F{alphameric} | TYPE | QUALITIES |

NAME | ADDRESS | PLACE | WHERE |

BACK | PREVIOUS | LINK |

DIRECTION | ORDER | WHEN | TIME |

SPEED | STATE | VALUE | COLOR |

RESOURCES | DOER | USER | START

expand Æ hop | move | create | linkup

transfer Æ run | call | output

modify Æ split | partition | select | replicate | integer

and | par and

advance Æ

advance | sync advance

repeat Æ cycle | loop | sling | repeat | repeat sync

echo Æ rake | min | max | sort | sum | average |

product | count | state

arithmeticÆ add | subtract | multiply | divide | degree

structuralÆ separate | unite | concatenate | append |

intersect | content | index | rand

assign Æ assign | assign peers

more equal | empty | nonempty |

belong | not belong | inside | not inside

timing Æ sleep | remain

type Ænodal | heritable | frontal | environmental |

info | matter | number | string | wave

PROVIDING SPATIAL INTEGRITY FOR DISTRIBUTED UNMANNED SYSTEMS

The DSL top level structure can also be expressed

graphically, as in Fig. 1. The basic construct, rule,

can represent any action or decision and can, for

example, be as follows:

 Elementary arithmetic, string or logic operation.

 Hop in a physical, virtual, or combined space.

 Hierarchical fusion and return of (remote) data.

 Distributed control, both sequential and parallel.

 A variety of special contexts for navigation in

space, influencing operations and decisions.

 Type or sense of a value, or its chosen usage,

guiding automatic interpretation.

constant

number

st rin g

sp ecial

identifier

reserved

wave

[ rule ] ( { wave , } )

variable

evolution

expansion

branching

advancing

repetition

granting

echoing

processing

combining

assignment

comparison

membership

linkage

type

usage

fusion verification essence

Figure 1: Recursive structure of DSL.

Different variants of this syntax and semantics

had been implemented for previous DSL subsets

(Sapaty, 1999, 2005), where conventional

expression of operations and delimiters between

program parts can be used too, say, for better

readability and compactness. For example:

add(3, 5, 7) same as 3 + 5 + 7

advance(w1, w2, w3) same as w1; w2; w3

where w1 to w3 may be arbitrary DSL programs

(waves) themselves. The first example could have

any programs instead of just numbers, each

returning its (possibly, remote and multiple) results,

as follows:

add(w1, w2, w3) same as w1 + w2 + w3

3 DISTRIBUTED INTERPRETER

A variety of options may be available for automatic

interpretation of DSL scenarios – from fully

centralized and sequential to fully distributed and

parallel. Due to peculiar syntax and semantics, the

language interpretation in distributed systems is

transparent and straightforward. Some basic features

of the DSL interpretation are as follows.

 Direct association of props with world points

drastically simplifies bringing data from the

points to scenarios or vice versa: scenarios or

their parts to world points.

 Chained actions can self-navigate and match the

world, while omitting used “heads” and

forwarding remaining “tails” further.

 Independent actions can be launched in parallel,

developing autonomously in parts of the world.

 The interpreter copy can be installed in internet

hosts, mobile robots, laptops, mobile phones,

smart sensors, or implanted into biological units.

 The interpreter can also be a human being,

performing manually of what is for herself while

passing other parts of the scenario to other

human or electronic interpreters and establishing

dynamic command and control infrastructures

between them.

 Any other systems can be accessed via the

networked interpreters, the latter forming a

supervisory layer managed in DSL.

 The interpreter copies may be concealed inside

the systems to be impacted, even without their

knowledge (to work in hostile environments).

 The interpreters can also migrate in the worlds to

be managed, collectively executing (mobile too)

mission scenarios, resulting altogether in a

flexible and ubiquitous system organization.

 The DSL interpreter consists of a number of

specialized modules working in parallel and

handling and sharing specific data structures,

which are supporting both persistent virtual

worlds and temporary hierarchical control

mechanisms (Sapaty, 1999, 2005).

 The heart of the distributed interpreter is its

spatial track system enabling hierarchical

command and control and remote data & code

access, also providing high integrity of emerging

parallel and distributed solutions, achievable

without central facilities.

In application to robotic communities, the

approach allows us to convert any group of mobile

robots into a goal-directed cooperative system by

integrating copies of the DSL interpreter, presented

as a universal control module U in Fig. 2, with

traditional robotic functionalities, as in

Kuhnert,

Krödel, 2005. (The figure exhibits mobile robots

which participated in M-ELROB 2008 trial).

Any mission scenario in DSL can start from any

robot, covering and tasking the whole system (or its

parts needed) at runtime and in parallel.

Subordination between the units and dynamic

command and control are established automatically,

as a derivative of the mission scenario and current

state of environment.

ICINCO 2009 - 6th International Conference on Informatics in Control, Automation and Robotics

Figure 2: Heterogeneous robotic teaming using embedded

DSL interpreters.

Due to fully interpretive nature of the technology,

the scenarios can self-recover from any points,

timely reacting on failures of robots. The whole

group may remain fully functional and global-goal-

oriented even in case of indiscriminate damages to

individual units.

4 ELEMENTARY EXAMPLE

An elementary task to be programmed in DSL may

look like follows:

Go to given physical locations of the disaster

zone (represented in a proper system of coordinates

by the three locations): (50.433, 30.633),

(50.417, 30.490), and (50.467, 30.517). Evaluate

damage in each location, then find and transmit the

maximum destruction value on all locations,

together with exact coordinates of the corresponding

location, to a management center.

The corresponding program in DSL will be:

transmit(maximum(

move((50.433, 30.633),

(50.417, 30.490),

(50.467, 30.517));

attach(assess(damage),WHERE)))

This program reflects semantics of the task to be

performed in a distributed space, regardless of

possible equipment that can be used for this. The

latter may, for example, be a set of sensors scattered

in advance throughout the disaster zone, where

hopping by coordinates may result in a wireless

access of the sensors already present there, not

necessarily moving into these points physically.

As another solution, the program may task

mobile robots to move into these locations and

perform the needed damage assessment upon

reaching the destinations. We will be showing here

this latter option, using three available robots R1, R2,

and R3.

The possible starting position and initial scenario

injection (let it be into R1) are shown in Fig. 3.

50.417, 30.490

50.433, 30.633

50 .467 , 3 0.517

transmit(maximum(

move((50.433,30.633),

(50.417,30.490),

(50.467,30.517));

attach(assess(damage),WHERE)))

Figure 3: Initial scenario injection.

After the creation of a distributed interpretation

infrastructure covering all three robots, R1 is

partitioning the scenario, and modifying and tasking

itself and the other two robots, as in Fig. 4.

move(50.467,30.517);

attach(assess(damage),

WHERE)

transmit

maximum

50.417, 30.490

50.433, 30.633

50.467, 30.517

DSL Interpretation

Infrastructure

move(50.467,30.517);

attach(assess(damage),

WHERE)

move(50.467,30.517);

attach(assess(damage),

WHERE)

Figure 4: Parallel tasking of three robots.

All three robots then move independently to the

locations optimally chosen for them, as in Fig. 5.

DSL Interpretation

Infrastructure

50.417, 30.490

50.433, 30.633

50.467, 30.517

transmit

maximum

move(50.467,30.517)

move(50.417,30.490)

move(50.433,30.633)

Figure 5: Simultaneous robot movement.

PROVIDING SPATIAL INTEGRITY FOR DISTRIBUTED UNMANNED SYSTEMS

In each location reached independently by a

corresponding robot, the damage assessment and

exact coordinates return and attachment take place,

as in Fig. 6.



transmit

maximum

50.417, 30.490

5 0. 4 33 , 3 0. 6 33

50.467, 30.517

asse ss

assess

O bser ved

damage

Obse rv ed

damage

Observed

da m a g e

Figure 6: Simultaneous damage assessment.

And finally, R1, using rule maximum, finds global

maximum damage value from those obtained in each

of the three robots, and together with the

corresponding location coordinates transmits it to

the management center, as in Fig. 7.



transmit

maximum

Dam age leve l &

co or di n at e s

50 .4 17, 30 .4 90

50.433, 30.633

50 .4 67, 30 .5 17

Damage level &

coordinates

DSL interpretation

infrastructure

Dam age leve l &

coor dinate s

Management

Ce nter

Figure 7: Merging data, finding global maximum.

As can be seen from the examples above, a semantic

level scenario describing what to do in the

distributed space, can be interpreted by robotic

teams autonomously, and by different numbers of

cooperating robots (we could use two or a single

robot instead). The number of available robots can

also vary at runtime, during the scenario evolution.

5 MORE ROBOTIC SCENARIOS

We will be using here the main scenarios that were

the basis of the M-ELROB 2008 trial.

5.1 Non-urban Reconnaissance

For this scenario, it is supposed that a group of

unknown vehicles is located in some distance in a

non-urban area (defined, say, with the position of a

center and area’s radius), with security situation

unclear there, so the reconnaissance should be done

by robotic vehicles for not risking own personnel.

The objective is to go to this target area and search

for vehicles with specific characteristics. If found,

they should be examined in detail, with their

parameters collected and reported to the control

station.

The general picture is shown in Fig. 8, where the

reconnaissance facilities should first go to the target

area (i.e. its center), observe the area by

cameras/sensors to roughly locate most probable

targets (by their size, for example). The next will be

to move directly to these selected targets and sense

& collect their detailed parameters, with sending the

results to the control point where they are stored and

analyzed.

Start

Target

area

Target

vehicles

Radius

Center

Find

Result 3

Move

Result 2

Result 1

Move

Figure 8: The reconnaissance scenario.

Parallel Solution. This, in DSL, may allow us to use

as many reconnaissance vehicles as possible (a

single one including), potentially involving

individual vehicles for each target identified, for

their detailed examination.

USER =(move(start); WHERE=center;

Targets=recognize(radius,features);

split(Targets); WHERE = VALUE;

collect(size, type, speed))

Explicitly Sequential Solution. The following DSL

program just details navigation and organization

procedures to execute the reconnaissance scenario in

a strictly sequential way, which may be useful for

optimization of the use of a single vehicle only.

move(start); WHERE = center;

Targets=recognize(radius,features);

loop(WHERE = withdraw(Targets,1);

Result &=collect(size,type,speed));

USER = Result

Avoiding Obstacles. The

movement to the target area and inside it may be

complicated due to presence of obstacles, as shown

in Fig. 9. The following DSL program, for the move

ICINCO 2009 - 6th International Conference on Informatics in Control, Automation and Robotics

from Start to Center, uses an external procedure

approach_or_stop to detect obstacles and stop to

avoid collision, and the procedure

suitable to find

next suitable waypoint on the way to the destination,

from which the move should continue.

move(start);

loop(approach_or_stop(center);

WHERE != center;

WHERE = suitable(depth,center));



Start

Target

area

Target

vehicles

Radius

Center

Find

Move

Obstacle

Stop

Figure 9: Avoiding obstacles.

5.2 Camp Security

For the camp security scenarios, a defined urban

area has to be monitored (think military camp) and

this should be executed by robotic vehicles too, to

minimize risk to human personnel. The objective is

to detect and report irregularities in the area, like

intruders, while acquiring their positions and

imagery, and transmitting to control station.



Robot 2

Robot 1

Rob o t 3

0_2

3_2

1_2

0_0

3_12_11_1

2_01_0

Figure 10: Camp security scenario.

The general picture is shown in Fig. 10, where

the camp units (numbered 1 to 6) are simultaneously

patrolled by a number of robotic vehicles moving

along the paths between and around the buildings.

Distributed Campus Map. The proper routing of

vehicles and resolution of possible conflicts between

them (like collision avoidance) can be assisted by

the creation of a distributed map of the campus area

(just reflecting Fig. 10) by the following DSL

program (with node names reflecting X-Y

coordinates of the crossings, and all links named r):

create(#3_1; F1=A; r#2_1; F2=A;

r#1_1; F3=A; r#0_1;(r#0_2; r#1_2;

r#F3, (r#2_2; r#F2,(r#3_2; r#F1))),

(r#0_0; r#1_0; r#F3, (r#2_0; r#F2,

(r#3_0; r#F1))))

Random Movement. The next program organizes the

duty performance by three parallel processes (which

may be executed by three robots) using the created

distributed map, with random choice of the next-hop

crossing and activation of the external service

procedure

move_check_report to analyze the

local security situation while on the move.

hop(0_1, 2_2, 3_0); WHERE = CONTENT;

repeat(or((hop(link(random));

grasp(Mark == nil; Mark = 1);

(hop(BACK); Mark) = nil;

move_check_report(CONTENT)), stay))

Movement via Predetermined Routes. If to use

predetermined routes only, like the ones shown in

Fig. 11 (one route using links named

r1 and another

one

r2), the collisions between robots can be

avoided in full.

546

Robot 1

Robot 2

0_2

3_2

1_2

0_0

3_1

2_11_1

2_01_0

2_2

Figure 11: Using predetermined routes.

Additional links r1 and r2 in the campus map can

be installed by the following DSL program:

Linkup((#0_2; r1#1_2; r1#1_1; r1#1_0;

r1#0_0; r1#0_1; r1#0_2),

(#3_2; r1#2_2; r1#2_1; r1#2_0;

r1#3_0; r1#3_1; r1#3_2))

And two independent spatial processes navigating

the campus via the new links (which may engage

two robots) can be organized by the following

parallel DSL code:

(hop(0_1); Flink = +r1),

(hop(3_0); Flink = +r2);

PROVIDING SPATIAL INTEGRITY FOR DISTRIBUTED UNMANNED SYSTEMS

WHERE = CONTENT;

repeat(hop(link(Flink));

move_check_report(CONTENT))

Any imaginable combinations of different types of

simultaneous movement through the camp (like

those by predetermined routes and/or by free,

random, wandering) with collision avoidance can

also be easily organized in DSL.

5.3 Transport Convoy

Imagine there is a delivery for a camp located in

some distance. The objective is to move at least two

vehicles to the target location, where only the first

one can be manned and the second should follow the

route of the first one, on a certain distance from it.

We will consider a fully robotic solution for such a

convoy, with two and also any number of vehicles,

where only the first vehicle knows (and follows)

waypoints toward the target location, while others

dynamically chaining with, and following the

previous ones on the move.

Two-unit Convoy. It is represented by the

communicating Leader and Follower, where the

first one defines its movement by a sequence of

waypoints, and the second one, regularly requesting

the Leader, moves to the positions previously

occupied by it, while keeping a certain threshold

distance. This is shown in Fig. 12, and by the DSL

program that follows.

Waypoints

Range

Leader

Follower

Start

Leader,

Follower

Figure 12: Two-unit convoy.

move(start);

(create(Leader);

Waypoints = (w1, w2, w3, …);

loop(WHERE = withdraw(Waypoints,1))),

(create(Follower); sling(

Lcoord = (hop(range, any); WHERE);

distance(WHERE, Lcoord) > threshold;

WHERE = Lcoord))

Multiple-unit Convoy. A scenario for the convoy

with any number of chained processes (to be

materialized by robotic units) is described by the

following DSL program and depicted in Fig. 13. For

this case, only the first process is a pure leader and

the last process a pure follower, while all other

processes combine both functionalities, i.e. being

followers for the previous processes and leaders for

the subsequent ones.

move(start);

cycle(N < number; create(N += 1));

(NAME == 1; Waypoints = (w1,w2,w3,…);

Loop(WHERE = withdraw(Waypoints,1))),

(NAME != 1; sling(

Lcoord = (hop(range,NAME-1); WHERE);

distance(WHERE, Lcoord) > threshold;

WHERE = Lcoord))

Waypoints

Range

Leader

Follower

Leader

Follower

Leader

Follo wer

Follower

Start

Nodes:

1, 2, 3, 4, 5

Figure 13: Multiple-unit convoy.

5.4 Transport Mule

Fir this scenario, there are two camps with a certain

distance in between, and a cargo with a given weight

should be transferred between the camps. We will

consider here different possibilities to deliver

payload between the camps, using unmanned

vehicles as “mules”.

In a Single Piece. This may be the case if cargo’s

weight allows it to be put on a single vehicle, as

shown ion Fig. 14.



Car go

Campus2Campus1

Figure 14: Single piece cargo delivery.



Limit

Campus2Ca m pu s1

Figure 15: Shuttling delivery.

The related DSL program will be as follows:

move(Campus1);

frontal(Cargo) = “substance”;

move(Campus2); Store = Cargo

Shuttling between Camps. For this option, the

process shuttles as often as possible between the two

camps after partitioning the cargo into portions for

the weight allowed, unless all the cargo is delivered,

ICINCO 2009 - 6th International Conference on Informatics in Control, Automation and Robotics

as shown in Fig. 15 and by the following program.

move(Campus1); frontal(Load);

Cargo = “substance”; Limit = 50;

loop(or((weight(Cargo) > Limit;

Load = withdraw(Cargo,Limit)),

(weight(Cargo) > 0; Load = Cargo));

hop(Campus2); Store += Load;

hop(Campus1))

Multiple, Parallel Delivery. For this case, different

processes (vehicles) are considered to be

independent from each other, each moving to the

destination as quickly as possible on its own (see Fig.

16 and the following program).

move(Campus1); frontal(Load);

Cargo = “substance”; Limit = 50;

cycle(or((weight(Cargo) > Limit;

Load = withdraw(Cargo, Limit)),

(weight(Cargo) > 0; Load = Cargo)));

move(Campus2); Store += Load

Limit

…

Campus2

Campus1

Figure 16: Parallel cargo delivery.

Multiple, Convoy Delivery. For this scenario, the

vehicles, each with a limited partition of cargo, are

dynamically chaining in a column for a cohesive

movement towards the destination (see Fig. 17 and

the subsequent DSL program).

move(Campus1); frontal(Load);

Cargo = “substance”; Limit = 50;

cycle(or((weight(Cargo) > Limit;

Load = withdraw(Cargo, Limit)),

(weight(Cargo) > 0; Load = Cargo));

create(N += 1));

(NAME == 1; move(Campus2)),

(NAME != 1; loop(WHERE != Campus2;

WHERE = (hop(NAME-1); WHERE));

Store += Load

Limit

…

Limit

Campus2

Campus1

Figure 17: Delivery in a convoy.

5.5 Explosive Ordnance Disposal

Explosive Ordnance Disposal (EOD) means the

detection, identification, onsite evaluation, rendering

safe, recovery, and final disposal of Unexploded

Ordnance (UXO) including detonation and burning.

It is often said that the EOD operation is a 3 Ds one,

which is Dangerous, Dirty and Demanding (or

Difficult) job. Using robotic vehicles, especially

multiple ones, is therefore becoming the most

promising EOD option.

Various kinds of EOD scenarios for navigation

and examination of the target territory may be

offered. We will just hint here on the simplest two

options, easily expressible in DSL.

Sequential Territory Search. This represents a

single-thread process (oriented on a single vehicle),

where the whole territory is incrementally scanned

unless all being searched, as described by the

following program and depicted in Fig. 18.

Start

Finish

Figure 18: Sequential navigation.

X1 =…; X2=…; Y = Y1 =…; Y2 =…; DY =…;

loop(WHERE = (X1,Y); WHERE = (X2,Y);

(Y += DY) < Y2; WHERE = (X2,Y);

WHERE = (X1,Y); (Y += DY) < Y2)

The sequential coverage of the territory can be

organized with minimum waypoints to pass, in a

zigzag way, as shown in Fig. 19 and by the

following program.

Start

Finish

Figure 19: Sequential zigzag navigation.

X1 =…; X2=…; Y = Y1 =…; Y2 =…; DY =…;

loop(WHERE = (X1,Y); (Y += DY) < Y2;

WHERE = (X2,Y); (Y += DY) < Y2)

Another solution, starting from the region’s

periphery and then gradually moving to its center, is

shown in Fig. 20, and by the next DSL program.

PROVIDING SPATIAL INTEGRITY FOR DISTRIBUTED UNMANNED SYSTEMS



Start

Fin ish

Figure 20: Sequential out-in navigation.

X1 = …, X2 = …; Y1 = …; Y2 = …;

DX = …; DY = …; X = X1; Y = Y1;

DDX = X2 – X1; DDY = Y2 – Y1; N = 1;

WHERE = (X,Y);

loop(Y += DDY * N; WHERE = (X,Y);

X += DDX * N; WHERE = (X,Y);

(DDX -= DX) > 0; (DDY -= DY) > 0;

N *= -1)

Parallel Territory Search. This

can be represented by a number of independent

processes, each starting from a different location,

and navigating altogether the whole region in

parallel, as depicted in Fig. 21, and explained by the

DSL program that follows (taking into account that

all processes follow predetermined routes for this

case).

X1 X2

Figure 21: Parallel predetermined navigation.

X1 =…; X2=…; Y1 =…; Y2 =…; DY =…;

frontal(Y) = Y1; DDY = 0;

cycle((Y += DDY) < Y2; DDY += DY);

WHERE = (X1,Y); WHERE = (X2,Y)

Parallel search of the territory can also be organized

in a random way, where each process randomly

chooses its next hop, also taking into account that

the chosen next destination should not have been

visited before (at least to look like this, with the help

of visual sensors). Parallel random search may have

an advantage before predetermined search in that it

can eventually cover all the territory despite possible

failures of individual processes (robots). Such a

search, with processes starting from some initial

points (named

c1 to c5), where processes also keep

certain threshold distance from each other, is shown

in Fig. 22 and by the following program.

Range

Random

New

Random

Figure 22: Parallel random navigation.

move(c1, c2, c3, c4, c5); Range = …;

X1 =…; X2=…; Y1 =…; Y2 =…; D =…;

loop(New = WHERE + 2(random(-D, D));

inside(New, (X1, X2, Y1, Y2));

hop(New, Range) & seen(New) == nil;

shift_check_act(New))

6 OTHER APPLICATIONS

Many other applications of the paradigm are

possible, as follows, some of which already

investigated, tested, and published (Sapaty, 1999,

2002, 2005, 2007, 2008, 2008a;

Sapaty, Sugisaka,

Finkelstein, Delgado-Frias, Mirenkov, 2006; Sapaty,

Morozov, Sugisaka, 2007

Emergency Management. Using interpreters

installed in massively wearable devices may allow

us to assemble workable systems from any wreckage

after the disasters, using any remaining

communication channels, manual including. These

emergent systems can provide distributed self-

awareness, collect statistics of casualties, guide the

delivery of relief goods, and coordinate collective

escape from the disaster zone.

Directed Energy Systems.

The technology can

provide high flexibility in organizing directed

energy (DE) systems, especially in crisis situations,

making automatic distributed decisions with the

“speed of light” too. It may also help automate the

global power dominance by optimized delivery of

directed energy into any world points via

dynamically organized networks of relay mirrors.

Distributed Avionics.

Implanting interpreter copies

into main control points of the aircraft may provide a

higher, intelligent, layer of its self-analysis and

recovery, by the spreading recursive scenarios

starting from any point and collecting & fusing key

data from other points. The embedded interpretation

network with local, dynamic, and emergent links

will be fully functional under any damages,

especially with wireless communications between

ICINCO 2009 - 6th International Conference on Informatics in Control, Automation and Robotics

the interpreters. This may always provide global

control integrity, even in a physically disintegrating

object, helping to save lives and complete missions.

Sensor Networks. Wireless sensors may be dropped

from the air massively, as “smart dust”. Having a

limited communication range, they must operate in a

network to do nonlocal jobs in a distributed

environment. With the technology offered, we can

convert their emergent networks into a universal

parallel computer operating in DSL. It can

effectively solve complex distributed problems--

from just collecting and fusing scattered data to

outlining and assembling images of the distributed

phenomena like, for example, flooding, smog, flocks

of birds, movement of troops, etc.

Advanced Command and Control. In DSL, it is

possible to define high-level scenarios concentrating

on mission goals and top decision-making while

delegating C2 routines, appearing at runtime as a

derivative of the mission and environment states, to

automatic interpretation. It is also convenient to

express in DSL any theoretical and practical issues

of advanced C2 explicitly.

Infrastructure Protection. Navigating the systems at

runtime, the technology can analyze safety and

integrity of critical infrastructures and key resources,

establishing protective networked mechanisms

throughout them. Other systems can be involved

from the WPT layer for emergent infrastructure

protection and recovery, including air and space

defense, police and army. In relation to energy

infrastructures, the technology can help observe

power networks from the air or ground, trace

electric, gas, or oil supply lines, sensing their states

(and, if needed, directly accessing the disaster

zones), also providing regular or emergent sentry

duties at power installations, etc.

Global and Battlespace Dominance. The DSL

scenarios, using any electronic media, can self-

spread, outline, and grasp distributed systems of

different natures while establishing global

dominance over them. They can analyze their

internal infrastructures, finding strong and weak

points, orient behavior, or destroy the infrastructures

or the system as a whole if required. The approach,

as an intelligent self-recovering super-virus, which is

difficult to discover and kill by traditional means,

can effectively employ advanced robotic facilities,

like swarms of aerial and ground vehicles, to attack

adversarial systems.

7 GESTALT-RELATED

Our approach may be considered as one of the first

attempts to formalize and implement the notion of

gestalt (Wertheimer, 1924), under which the whole

dominates over parts (being greater than the sum of

them), with parts having sense only in the context of

the whole, rather than vice versa. Gestalt theory

represented the main departure from atomistic vision

of systems at the beginning of the last century. Many

existing systems, especially distributed ones, are still

based on the concept of predetermined parts (agents)

that communicate with each other in an attempt to

get the global behavior needed. The latter, with

rapidly growing number of agents in complex

systems and starting from the agents level is

becoming more and more problematic.

Within the approach offered, we have come to

quite a different and higher level model of the

system organization. Abstracting from system parts

and their interactions, which may be emergent and

varying at runtime, we can describe the needed

global system behavior on a semantic level, where

parts and their interactions may not be known in

advance, and may dynamically appear (disappear

too) just to maintain the global behavior needed.

The technology developed allows us to

automatically interpret global system scenarios in

any networked systems (comprising internet hosts,

laptops, mobile robots, mobile phones, smart sensors,

and/or humans themselves). It allows us to get even

higher—to describe what the system should do on

the highest level, where its local and global behavior

is a derivative of this description, which, in its turn,

makes the system structures and operation as a

further derivative.

8 CONCLUSIONS

A novel ideology and technology, converting any

distributed system into a universal spatial machine

capable of solving complex problems on itself and

on the surrounding environments, has been

presented. This conversion can be achieved by

implanting into the system sensitive points and its

active doers, humans and robots including, of the

same copy of a universal control module,

communicating with other such modules via

available channels. Their entire network, which may

be dynamic and emergent, collectively interprets

mission scenarios written in a special high-level

language, which are defining system’s internal and

external behavior.

PROVIDING SPATIAL INTEGRITY FOR DISTRIBUTED UNMANNED SYSTEMS

Created and modified on the fly, the scenarios

can start from any component, covering the system

at runtime through the cooperating interpreters.

During the scenario evolution, any operations can be

carried out throughout the distributed world, along

with the needed movement of code, equipment and

artificial or biological doers, humans including, as

well as creation and maintenance of physical and

virtual infrastructures supporting the missions.

The approach offered can dramatically simplify

application programming in distributed systems,

especially robotized ones. As can be seen from the

examples throughout this paper, programming multi-

robot scenarios in distributed and dynamic

environments in DSL may not be more difficult than,

say, programming of routine data processing tasks in

traditional languages like Fortran, C, or Java.

The distributed robotized systems are of rapidly

growing importance in many areas, and especially in

defense, where robotic swarming on asymmetric

battlefields is becoming a major dimension of the

new military doctrine for the 21

century (Singer,

2009). The written above is much in line with these

trends, allowing us to flexibly combine loose

swarming with more classical command and control,

which can help gradually transform fully manned

into mixed and ultimately totally unmanned systems.

Other prospective applications of this work can

be linked with economy, ecology and weather

prediction—by using the whole networked world as

a spatial supercomputer, self-optimizing its

performance.

The approach offered may also be compared

with the invention of the first world computers and

first high-level programming languages (Zuse,

1948/49; Rojas, 1997). In our case, this computer

may not only operate with data stored in a localized

memory, but can cover, grasp, and manage any

distributed systems, the whole world including, and

can work not only with information but with

physical matter or physical objects too.

ACKNOWLEDGEMENTS

This work has been funded by the Alexander von

Humboldt (AvH) Foundation in Germany. Special

thanks to Klaus-Dieter Kuhnert and Matthias Langer

for sheltering this activity at the University of

Siegen. Years of cooperation with Robert

Finkelstein (Robotic Technology Inc., USA) and

Masanori Sugisaka (Nippon Bunri University,

Japan) contributed much to the ideas expressed in

this paper. The support of this ideology and

technology by Joaquim Filipe (Escola Superior de

Tecnologia, Portugal) and by ICINCO conferences

was really invaluable. Encouragement from Stephen

Lambacher (Aoyama Gakuin University, Japan) has

been appreciated too.

REFERENCES

Kuhnert, K.-D., Krödel, M., 2005. Autonomous Vehicle

Steering Based on Evaluative Feedback by

Reinforcement Learning. MLDM.

M-ELROB, 2008. Military European Land-Robot Trial.

Hammelburg, Germany.

Rojas, R., 1997. Konrad Zuse’s Legacy: The Architecture

of the Z1 and Z3. IEEE Annals of the History of

Computing. Vol. 19, No. 2.

Sapaty, P. S., 1999. Mobile Processing in Distributed and

Open Environments, John Wiley & Sons. New York.

Sapaty, P. S., 2002. Over-Operability in Distributed

Simulation and Control. The MSIAC's M&S Journal

Online. Winter Issue, Volume 4, No. 2, VA, USA.

Sapaty, P. S., 2005. Ruling Distributed Dynamic Worlds,

John Wiley & Sons. New York.

Sapaty, P., Sugisaka, M., Finkelstein, R., Delgado-Frias, J.,

Mirenkov, N., 2006. Advanced IT Support of Crisis

Relief Missions. Journal of Emergency Management,

Vol. 4, No. 4.

Sapaty, P., Morozov, A., Sugisaka, M., 2007. DEW in a

Network Enabled Environment. Proc. International

conference Directed Energy Weapons 2007. Le

Meridien Piccadilly, London, UK.

Sapaty, P., 2007. Intelligent management of distributed

sensor networks, In Sensors, and Command, Control,

Communications, and Intelligence (C3I) Technologies

for Homeland Security and Homeland Defense VI, ed.

by E. M. Carapezza. Proc. of SPIE Vol. 6538, 653812.

Sapaty, P., 2008. Distributed Technology for Global

Dominance. Proc. of SPIE, Volume 6981, Defense

Transformation and Net-Centric Systems 2008. Raja

Suresh, Ed., 69810T.

Sapaty, P., 2008a. Grasping the Whole by Spatial

Intelligence: A Higher Level for Distributed Avionics.

Proc. International Conference Military Avionics

2008. Cafe Royal, London, UK.

Sapaty, P., 2009. Gestalt-Based Ideology and Technology

for Spatial Control of Distributed Dynamic Systems.

Proc. International Gestalt Theory Congress, 16th

Scientific Convention of the GTA. University of

Osnabrück, Germany.

Singer, P. W., 2009. Wired for War: The Robotics

Revolution and Conflict in the 21

Century, Penguin.

Wertheimer, M., 1924. Gestalt Theory, Erlangen. Berlin,

1925.

Zuse, K., 1948/49. “Uber den Plankalk, als Mittel zur

Formulierung schematisch kombinativer Aufgaben”,

In Archiv Mathematik, Band I.

ICINCO 2009 - 6th International Conference on Informatics in Control, Automation and Robotics