1 / 85

THESIS COLLOQUIUM - PowerPoint PPT Presentation

  • Uploaded on

THESIS COLLOQUIUM. Collision avoidance and coalition formation of multiple unmanned aerial vehicles in high density traffic environments. Joel George M.

I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
Download Presentation

PowerPoint Slideshow about ' THESIS COLLOQUIUM' - nitza

An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.

- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript


Collision avoidance and coalition formation of multiple unmanned aerial vehicles in high density traffic environments

Joel George M

“… it was nevertheless - the first time in the history of the world in which a machine carrying a man had raised itself by its own power into the air in full flight, had sailed forward without reduction of speed, and had finally landed at a point as high as that from which it started.”

Orville Wright

Details of first flight:

Speed = 6.8 miles/hour

Range = 120 feet

Altitude = 10 feet

Slogan of aircraft design industry of

Faster, Farther, Higher (and Safer)

Boundaries of speed, altitude, range, and endurance have been pushed

further and further

Aircraft kept the tag “machine carrying a man” of

Presence of man in aircraft was always an important design consideration

“Elimination of pilot from a manned combat of aircraft removes many of the conventional design constraints …

This at once throws open the design parameter space and dramatic improvements in performance measures like increased speed, range, maneuverability, and payload can be achieved.”

Late Dr. S Pradeep

Why Unmanned Aerial Vehicles (UAVs)? of

In some missions, human presence ‘need not’ be there

In some other missions, human presence ‘should not’ be there

Unmanned Aerial Vehicles find applications in

Dull, Dirty, and Dangerous missions

Why UAVs? of

Factors compelling the use of Unmanned Aerial Vehicles (UAVs)

Design freedom (mission specific designs)

Dull, dirty, and dangerous missions

Low cost, portability, absence to human risk, …

Why autonomous UAVs? of

UAVs can be remotely piloted

However, desirable to make UAVs autonomous

Why multiple UAVs? of

UAVs are often small

Some missions are more effectively done by multiple UAVs

Use of multiple UAVs leads to coordination problems

Collision avoidance, coalition formation, formation flying, …

This thesis addresses the problems of of

Collision avoidance,

Coalition formation, and

Mission involving collision avoidance and coalition formation

of multiple UAVs in high density traffic environments





Collision avoidance among multiple UAVs


Collision avoidance with realistic UAV models


Coalition formation with global communication


Coalition formation with limited communication


Coalition formation and collision avoidance in multiple UAV missions





Collision avoidance of

Using information of positions and velocities of UAVs in the sensor range, a UAV needs to find an efficient safe path to destination

A safe path means that no UAV should come within each others safety zones during any time of flight

Efficiency  less deviation from nominal path

Collision avoidance literature of

Have been looked at from the robotics and air traffic management points of view

Ground based robots can stop to finish the calculations

Collision avoidance algorithms addressing air traffic management problems consider only a few aircraft

Coalition formation of

Multiple UAVs with limited sensor ranges search for targets

A target found needs to be prosecuted

A UAV that detected the target may not have sufficient resources

‘Need to talk’ to other UAVs to form a coalition for target prosecution

Objective: To find and prosecute all targets as quickly as possible

The algorithm should be scalable

Coalition formation literature of

  • Multi-agent coalition formation

    • Can share resources

    • Extensive communication

  • Multi-robot coalition formation

    • Resources do not deplete

  • Multi-UAV coalition formation

    • Resources deplete with use

    • Need quick coalition formation algorithms

Collision avoidance and coalition formation in multiple UAV missions

Multi-UAV rendezvous with collision avoidance

Coalition formation with collision avoidance

CHAPTER 2 missions

Collision avoidance among multiple UAVs

Assumptions missions

UAV kinematic model

Constant speed

Minimum radius of turn

Further assumption

Limited sensor range

It suffices, in case of a multiple UAV conflict, missionsfor a UAV to avoid the most imminent near miss to obtain a good collision avoidance performance.

Two UAVs within each others safety missionszones results in a ‘near miss’

Objective is to reduce the number of near misses, as in a high density traffic case, it may not be possible to avoid near misses

Lesser the number of near misses, better the collision avoidance algorithm

Aircraft deviates from its nominal path due to collision avoidance maneuver.

Efficiency =

Lesser the deviation (higher efficiency), better the collision avoidance algorithm

UAVs encounter multiple conflicts missions

Reduce multiple conflicts to an ‘effective’ one-one conflict by finding

the ‘most threatening’ UAV from among the ones in sensor range

Most threatening UAV:

A UAV U2 is the most threatening UAV for U1 at an instant of time, if

U2 is in the sensor range of U1

Predicted miss distance between U1 and U2 suggests the occurrence of a near miss

Out of all the UAVs in the sensor range of U1 with which U1 has a predicted near miss, the near miss with U2 is the earliest to occur

Collision avoidance maneuver missions

A necessary condition for collision between two aircraft to occur is that

the Line of Sight (LOS) Rate between them be zero

For collision avoidance, a UAV does a maneuver to increase the LOS rate

Each UAV does a maneuver to avoid collision with the most threatening neighbor

Two Dimensional Reactive Collision missionsAvoidance: RCA-2D

Random flight test missions

Aircraft fly from random points on outer circle to

random points on inner circle

Velocity: 500 miles per hour

Turn rate: 5 degrees per second

Radius of outer circle 120 miles

Radius of inner circle 100 miles

Archibald, J. K., Hill, J. C., missionsJepsen, N. A., Strirling, W. C., & Frost, R. L. (2008). A satisficingapproach to aircraft conflict resolution. IEEE Transactions on System, Man, and Cybernetics - Part C: Applications and Reviews, 38, 510–521.

Since the test case involves random flights, the simulations are run 20 times for each case, and the values presented are averaged over the results obtained from these runs

Three dimensional engagement missions

Collision plane


Three dimensional collision avoidance algorithms


Modified random flights (three dimensional) random flights

Case 1: h = 20 miles,

rin= 100 miles, and rout = 120 miles

Case 2: h = 60 miles,

rin= 55 miles, and rout = 70 miles

Case 3: h = 100 miles,

rin= 40 miles, and rout = 50 miles

Summary of Chapter 2 random flights

Developed conceptually simple collision avoidance algorithms

For two and three dimensional conflicts

For high density traffic environments

Analyzed the performance of these algorithms

CHAPTER 3 random flights

Collision avoidance with realistic UAV models

Realistic random flights UAV Model

UAV of span 1.4224 m, weighing 1.56 kg

  • Stability and control derivatives from Aviones

  • A UAV flight simulator developed by the Brigham Young University

  • (an open source software)

  • Available:

The Zagi Aircraft

Span = 1.5 m

Mean Chord = 0.33 m

Weight = 1.5 kg

Picture courtesy:

UAV control system random flights

Controllers designed through successive loop closure

Separate controllers for holding altitude, attitude, and speed

PI controllers with parameters tuned manually

Controller design random flights

Altitude hold controller

Similar controllers for attitude and speed holds are designed

Implementing the guidance commands random flights

Look-up graph for bank angle that will provide required turn rate

Test of collision avoidance random flights

A example of collision avoidance of 5 UAVs

The test case is set-upsuch that the avoidance of one conflict will lead into another

Test case of random flights for dense traffic random flights

Random flights test case

inner circle radius 400 m

outer circle radius 500 m

velocity 12 m/s

maximum turn rate 10 deg/sec.

Any approach of two UAVs within 10 m is considered a near miss

An approach within 2 m is a collision.

Implementation of 3 random flights D collision avoidance algorithm

Realization of pitch and turn rate commands

Results of the random flight test random flights case

for homogeneous UAVs

for heterogeneous UAVs

Summary of Chapter 3 random flights

Implemented collision avoidance algorithms on 6 DoF UAV models

Simulations with heterogeneous and non-cooperating UAVs

CHAPTER 4 random flights

Coalition formation with global communication

Coalition formation for search and prosecute mission random flights

Search targets and destroy them

The targets may have different requirements


  • Destroy the target is minimum time

  • Coalition should have minimum number of UAVs

  • Rendezvous on target to inflict maximum damage

Assumptions random flights

Limited sensor radius

Target locations are not know a priori

Limited resources that deplete with use

Stationary targets

Global communication

Theorem random flights : The minimum time minimum member coalition formation

for a single target is NP-hard

UAV that detects the target – Coalition leader random flights

Coalition leader initiates the coalition formation process

Deadlocks are handled by rules/protocols

Communication random flights protocol for coalition formation process

Two stage algorithm for coalition formation random flights

Stage I

Find a minimum time coalition

Stage II

Find a minimum member coalition

Stage I: Minimum time coalition random flights

Theorem: Finding minimum member coalition is NP-hard

Recruit members to coalition in the ascending order of their ETA to target

Theorem random flights : Stage I gives a minimum time coalition

Theorem: Stage I has polynomial time complexity

Stage II random flights

‘Prune’ the coalition formed in stage I to form a reducedmember coalition

Solution using Particle Swarm Optimization (PSO) random flights

Global solution of the search and prosecute problem using PSO

Target locations known a priori

Comparison of solutions random flights

Summary of Chapter 4 random flights

Coalition formation algorithm for search and prosecute mission

Two stage polynomial time algorithm

Efficacy of the algorithm demonstrated via simulations

CHAPTER 5 random flights

Coalition formation with limited communication

UAVs have limited communication ranges random flights

Dynamic network over which coalition formations should take place

Network properties random flights

Every UAV acts as a relay node

Each hop of message has an associated lag

Log of messages kept to avoid duplication

Time-to-live for a message

Coalition formation over dynamic network random flights

Find a sub static coalition formation period

A UAV accepts to be a relay node only if sub-network that is over the UAV it is in communication range for the entire coalition formation period

Works well as coalition formation period is much shorter than the time scale in which network connection varies

Prosecution random flights sequence for maneuvering target

Rendezvous at a maneuvering target is difficult  sequential prosecution

Coalition leader tracks the maneuvering target and broadcast this information until the target is in the sensor range of one of the coalition members

A coalition member prosecutes the target and continues to track it until the target is within the sensor range of the next coalition member

Example random flights

Summary of Chapter 5 in communication delay

Coalition formation of UAVs with limited communication ranges

Prosecution of stationary, constant velocity, and maneuvering targets

CHAPTER 6 in communication delay

Collision avoidance and coalition formation in multiple UAV missions

Rendezvous – meeting at a pre-planned time and place in communication delay

Rendezvous of multiple UAVs

For simultaneous deployment of resources

To exchange resources or critical information

Multiple UAV Rendezvous in communication delay

Rendezvous under collision avoidance

Rendezvous of multiple UAVs when some of the UAVs have to do

collision avoidance maneuvers en route

Uses a consensus on Estimated Time of Arrival (ETA) at target

Multiple UAV Rendezvous Algorithm in communication delay

Consensus in ETA achieved using

Velocity control within bounds

Wandering maneuver

Change in velocity proportional to

(average ETA – ETA)

If velocity hits lower bound, then ‘wander away’ from the rendezvous point

Solution Approach in communication delay

Linear average consensus

  • In principle, any consensus protocol can be used.

  • Average consensus protocol is used for the purpose of illustration

Rendezvous: Simulation Results in communication delay

Rendezvous of 5 UAVs (3 of them do collision avoidance on the way)

Target tracking in communication delay

Summary of Chapter 6 in communication delay

Multiple UAV rendezvous with collision avoidance

Coalition formation with collision avoidance

CHAPTER 7 in communication delay


Algorithms for collision avoidance and coalition formation and their applications

Algorithms are

conceptually ‘simple’


Possible extensions of present work and their

Better controller implementations possible for collision avoidance

Better communication protocols possible for coalition formation