**7. Approach to the maneuver optimization**

One of the approach for optimizing the autonomous vehicle maneuver in the conditions of the opposition forces is to utilize advantages of modelling combat activities. During the military decision making process, commander and his staff members use a series of models and programs to accelerate and refine decision-making and operation missions. Example of good practise is exploitation of model of cooperative

**23**

UAS in the fleet.

*Military Factors Influencing Path Planning DOI: http://dx.doi.org/10.5772/intechopen.86421*

reconnaissance as part of Tactical Decision Support System [7]. According to [8], the fundamental approach is based on a sequence of procedures and the weighted integration of discrete layers, where all phases converge to a maneuver optimization issued from modified versions of Floyd-Warshall algorithm. Initial C++ application was designed for a basic experiments, providing relatively fast solution (derived from path-finding algorithms used in autonomous systems), whose task was to verify theoretical approach and the time profile of solution. If it is necessary, applica-

Fundamental theoretical approach in that case was inspired by Floyd-Warshall algorithm [8]. Original algorithm was pushed throughout several modifications that make it computationally applicable even for a large data structures comprising more than 106 nodes. Basic adjustment lay in elimination of so-called reverse cycles by

This process is carried out through the main field elements chosen for next phase solution. The status and verification cycle matches the bit position in the bit field with the position of the active element in the default structure. If the element belongs to a root what was modified (attribute is present), the element is excluded from the processing in the following iterative phase because it will be modified in

This process is theoretically simple; however, the realization of this step is relatively difficult in practice, because the memory performance **P**n achieves (2):

It means that models which contain more than 106 nodes must allocate over 125 GB memory only for genetic structure of each element. In the case of information transfer into other elements the amount of operations raise to a level that is incompatible neither with the real time application, nor on the fastest nowadays computers. It is therefore necessary to address the sub-problem in a different way and optimize the whole process by other approach. Previously mentioned idea

For the purpose of utilizing autonomous vehicle in role as reconnaissance means is the part of MDMP to plan the reconnaissance operation of the area of interest. For optimization of using more than one air autonomous vehicle so called as Unmanned Air Vehicle (UAV) in one assigned mission was developed model for optimization of

The objective of the model is specified in Formula (3) which is to minimize the maximum time of flight of all routes of individual UAS in the fleet. The details of the model along with its complete mathematical formulation can be found in [8].

minimize (max(*T*1,*T*2,…,*TN*)) (3)

where *Ti* is the time of flight of *i*-th UAS in the fleet, *N* is the total number of

The model of UAS Reconnaissance is similar to the Multi-Depot Vehicle Routing Problem (MDVRP) [9, 10]. The MDVRP is also about optimization of routes of a set of vehicles originating from multiple depots to visit a number of customers to deliver goods or services. There is, however, the significant difference between the models. While the main objective of the MDVRP is to minimize the total distance travelled by all vehicles which is expressed in Formula (4), the objective of the UAS Reconnaissance is to minimize the time of whole operation as already mentioned in Formula (1).

<sup>8</sup> (2)

tion could find alternative routes with more-favorable movement factor.

stopping the 82 calculation on all nodes in its root.

**<sup>P</sup>**<sup>n</sup> <sup>=</sup> **<sup>N</sup><sup>2</sup>** \_\_\_

where: N is the number of nodes (elements) of the graph.

works well but for a wide set of nodes (more than 106) is ineffective.

using swarm of UAV to effectively reconnaissance the area of interest [2].

the next steps.

#### *Military Factors Influencing Path Planning DOI: http://dx.doi.org/10.5772/intechopen.86421*

*Path Planning for Autonomous Vehicles - Ensuring Reliable Driverless Navigation...*

Preparation of autonomous vehicle to be put in place (time

Complexity/ease of operation of autonomous vehicle in all phases of task fulfillment (especially in the semi-automatic

Ability to select the axis of movement without detection by

Autonomous vehicle ability to overtake the enemy in its destruction (to detect and destroy the enemy before he

Heavy detection of autonomous vehicle by enemy means (autonomous vehicle does not produce sounds, smoke, other manifestations, minimizes radiation, etc.)

Inapplicability of autonomous vehicle to cooperate with

Ability to transmit information in the required ways x The existence of enemy means in the transfer space, which

Easily detection of autonomous vehicle by enemy means x

Sufficient capacity of source x Functionality of all components (sensors) x Interaction with "control center" at declared distance x Terrain Pass Through pass through x

autonomous vehicle controlled by the operator)

required to get ready for task start)

can destroy the autonomous vehicle

other vehicles in mass deployment

*\*Specified by the "assignor of the mission".*

the enemy

does so)

Etc.

**Table 1.**

**Factors Influence Effect weight** 

partially through

impassable x

x

x

x

**[number/**

**order]\* Positive Negative**

x

x

x

x

example, coefficients ranging from 1 to 10, a group of minor factors, coefficients in the range 1–6, and a group of least significant factors, coefficients in the range 1–3. The result of calculating the probability of completing a task is either left in a fraction that is adjusted to the "number" on the numerator or denominator side (e.g., 1/3.8 or 2.6/1). Such a fraction shape gives a multiple predominance of the

The possible factors affecting the ground autonomous vehicle (as listed in **Table 1**) to which the "assignor of a particular task" has to assign a specific value

One of the approach for optimizing the autonomous vehicle maneuver in the conditions of the opposition forces is to utilize advantages of modelling combat activities. During the military decision making process, commander and his staff members use a series of models and programs to accelerate and refine decision-making and operation missions. Example of good practise is exploitation of model of cooperative

probability of fulfilling/not fulfilling the task.

**7. Approach to the maneuver optimization**

before commencing (in the mission planning process).

*Example of evaluation factors influencing autonomous vehicle movement.*

**22**

reconnaissance as part of Tactical Decision Support System [7]. According to [8], the fundamental approach is based on a sequence of procedures and the weighted integration of discrete layers, where all phases converge to a maneuver optimization issued from modified versions of Floyd-Warshall algorithm. Initial C++ application was designed for a basic experiments, providing relatively fast solution (derived from path-finding algorithms used in autonomous systems), whose task was to verify theoretical approach and the time profile of solution. If it is necessary, application could find alternative routes with more-favorable movement factor.

Fundamental theoretical approach in that case was inspired by Floyd-Warshall algorithm [8]. Original algorithm was pushed throughout several modifications that make it computationally applicable even for a large data structures comprising more than 106 nodes. Basic adjustment lay in elimination of so-called reverse cycles by stopping the 82 calculation on all nodes in its root.

This process is carried out through the main field elements chosen for next phase solution. The status and verification cycle matches the bit position in the bit field with the position of the active element in the default structure. If the element belongs to a root what was modified (attribute is present), the element is excluded from the processing in the following iterative phase because it will be modified in the next steps.

This process is theoretically simple; however, the realization of this step is relatively difficult in practice, because the memory performance **P**n achieves (2):

$$\mathbf{P}\mathbf{n} = \frac{\mathbf{N}^2}{8} \tag{2}$$

where: N is the number of nodes (elements) of the graph.

It means that models which contain more than 106 nodes must allocate over 125 GB memory only for genetic structure of each element. In the case of information transfer into other elements the amount of operations raise to a level that is incompatible neither with the real time application, nor on the fastest nowadays computers. It is therefore necessary to address the sub-problem in a different way and optimize the whole process by other approach. Previously mentioned idea works well but for a wide set of nodes (more than 106) is ineffective.

For the purpose of utilizing autonomous vehicle in role as reconnaissance means is the part of MDMP to plan the reconnaissance operation of the area of interest. For optimization of using more than one air autonomous vehicle so called as Unmanned Air Vehicle (UAV) in one assigned mission was developed model for optimization of using swarm of UAV to effectively reconnaissance the area of interest [2].

The objective of the model is specified in Formula (3) which is to minimize the maximum time of flight of all routes of individual UAS in the fleet. The details of the model along with its complete mathematical formulation can be found in [8].

$$\text{minimize } \left\{ \max \left( T\_1, T\_2, \dots, T\_N \right) \right\} \tag{3}$$

where *Ti* is the time of flight of *i*-th UAS in the fleet, *N* is the total number of UAS in the fleet.

The model of UAS Reconnaissance is similar to the Multi-Depot Vehicle Routing Problem (MDVRP) [9, 10]. The MDVRP is also about optimization of routes of a set of vehicles originating from multiple depots to visit a number of customers to deliver goods or services. There is, however, the significant difference between the models. While the main objective of the MDVRP is to minimize the total distance travelled by all vehicles which is expressed in Formula (4), the objective of the UAS Reconnaissance is to minimize the time of whole operation as already mentioned in Formula (1).

*Path Planning for Autonomous Vehicles - Ensuring Reliable Driverless Navigation...*

$$\text{minimize} \left(\sum\_{i=1}^{N} Di\right) \tag{4}$$

where *Di* is the distance travelled by *i*-th vehicle, *N* is the total number of vehicles.

For verification of the proposed UAS Reconnaissance model there were designed two scenarios of tactical situation and applied experiments in real terrain with real UAV's [2, 11].
