Mission Planning for Unmanned Aircraft with Genetic Algorithms

Publication: ResearchPh.D. thesis

Abstract

Unmanned aircraft invokes different feelings in people. Some see ruthless killing machines, other see a potential for fast and cheap distribution of goods, yet other see flexible and convenient emergency rescue drones. Regardless, advances and miniaturization in motors, sensors, and computer processing power have taken the unmanned aircraft from being a military application to the commercial sector and even into the hands of hobbyists.
Still, the enthusiastic interest in the new technology and its prospective advantages overshadows the fact that it mainly sees application where the aircraft are mostly under human command, just like remote controlled planes have been for years. Actually the revolution of the drones is not so much a revolution of the unmanned aircraft as it is a digital control revolution. Only a few years ago, hopeful remote-control pilots had to invest countless hours of training in mastering the planes, the controls were complex and originated from the stick-and-throttle controls of real fighter airplanes. Now, inherently
unstable quad-copters can be controlled with touch screen, where a sliding motion to the right on the screen moves the aircraft to the right. Exactly such underlying automatic control methods has played a big role in popularizing the quad-copter as a toy, which in turn has awakened people’s imagination and enthusiasm.
The next step of the unmanned aircraft is to become fully autonomous. Expert
operators use unmanned aircraft to perform aerial surveys of nature conservation areas, construction sites, and the like. Although it flies autonomously, the operator need to understand the tool that the aircraft is to him. He must set the coordinates and altitude of each waypoint that the aircraft must visit, and he must decide on the sequence by which the waypoints must be visited. Surely, computer programs assist the operator in the planning process, but actually, the end product of the survey is not the flight plan of the aircraft or the aerial images that it takes, rather, it is the results of the analysis
of the images; the politician or the entrepreneur who ordered the analysis probably did not care how the data was collected. Just like the control algorithms in the autopilot relieved the operator of the piloting burden, the next step will relieve him of the planning burden. The analyst simply defines which analysis she wants to perform and a plan isautomatically created for the aircraft, which collects the needed data in the best possible fashion. All she has to do is release the aircraft and collect it again when it lands.
In this dissertation we study the automatic mission planning for unmanned aircraft.
The basis for the research is the case of agriculture automation where unmanned aircraft are used for aerial surveying of the crops. The farmer takes the role of the analyst above, who does not necessarily have any specific interest in remote controlled aircraft but needs the outcome of the survey. The recurring method in the study is the genetic algorithm; a flexible optimization framework that is used to perfect the flight plans.
Focus is given to planning under the kinematic constraints of the aircraft to obtain smooth trajectories that are much closer to a real flyable trajectory than the point-topoint waypoint trajectory. This focus results in the development of a method which models the aircraft as a Dubins vehicle and produces a plan that automatically decides on the headings and target speeds of a set of waypoints.
Another point of study is the constraint given by fuel limits. An aircraft can only
visit so many waypoints before it must refuel. A method is developed, which plans for refueling stops in the sequence of waypoints, so that the unmanned aircraft can continuously survey a given area. This area is an important direction for research into long-term autonomy, where robots work for hours or days without human intervention.
Two more technical contributions are made in the area of the genetic algorithms.
One is a method to decide on the right time to stop the computation of the plan, when the right balance is stricken between using the time planning and using the time flying.
The other contribution is a characterization of the evolutionary operators used in the genetic algorithm. The result is a measure based on entropy to evaluate and control the diversity of the population of the genetic algorithm, which is an important factor its effectiveness.
Close

Details

Unmanned aircraft invokes different feelings in people. Some see ruthless killing machines, other see a potential for fast and cheap distribution of goods, yet other see flexible and convenient emergency rescue drones. Regardless, advances and miniaturization in motors, sensors, and computer processing power have taken the unmanned aircraft from being a military application to the commercial sector and even into the hands of hobbyists.
Still, the enthusiastic interest in the new technology and its prospective advantages overshadows the fact that it mainly sees application where the aircraft are mostly under human command, just like remote controlled planes have been for years. Actually the revolution of the drones is not so much a revolution of the unmanned aircraft as it is a digital control revolution. Only a few years ago, hopeful remote-control pilots had to invest countless hours of training in mastering the planes, the controls were complex and originated from the stick-and-throttle controls of real fighter airplanes. Now, inherently
unstable quad-copters can be controlled with touch screen, where a sliding motion to the right on the screen moves the aircraft to the right. Exactly such underlying automatic control methods has played a big role in popularizing the quad-copter as a toy, which in turn has awakened people’s imagination and enthusiasm.
The next step of the unmanned aircraft is to become fully autonomous. Expert
operators use unmanned aircraft to perform aerial surveys of nature conservation areas, construction sites, and the like. Although it flies autonomously, the operator need to understand the tool that the aircraft is to him. He must set the coordinates and altitude of each waypoint that the aircraft must visit, and he must decide on the sequence by which the waypoints must be visited. Surely, computer programs assist the operator in the planning process, but actually, the end product of the survey is not the flight plan of the aircraft or the aerial images that it takes, rather, it is the results of the analysis
of the images; the politician or the entrepreneur who ordered the analysis probably did not care how the data was collected. Just like the control algorithms in the autopilot relieved the operator of the piloting burden, the next step will relieve him of the planning burden. The analyst simply defines which analysis she wants to perform and a plan isautomatically created for the aircraft, which collects the needed data in the best possible fashion. All she has to do is release the aircraft and collect it again when it lands.
In this dissertation we study the automatic mission planning for unmanned aircraft.
The basis for the research is the case of agriculture automation where unmanned aircraft are used for aerial surveying of the crops. The farmer takes the role of the analyst above, who does not necessarily have any specific interest in remote controlled aircraft but needs the outcome of the survey. The recurring method in the study is the genetic algorithm; a flexible optimization framework that is used to perfect the flight plans.
Focus is given to planning under the kinematic constraints of the aircraft to obtain smooth trajectories that are much closer to a real flyable trajectory than the point-topoint waypoint trajectory. This focus results in the development of a method which models the aircraft as a Dubins vehicle and produces a plan that automatically decides on the headings and target speeds of a set of waypoints.
Another point of study is the constraint given by fuel limits. An aircraft can only
visit so many waypoints before it must refuel. A method is developed, which plans for refueling stops in the sequence of waypoints, so that the unmanned aircraft can continuously survey a given area. This area is an important direction for research into long-term autonomy, where robots work for hours or days without human intervention.
Two more technical contributions are made in the area of the genetic algorithms.
One is a method to decide on the right time to stop the computation of the plan, when the right balance is stricken between using the time planning and using the time flying.
The other contribution is a characterization of the evolutionary operators used in the genetic algorithm. The result is a measure based on entropy to evaluate and control the diversity of the population of the genetic algorithm, which is an important factor its effectiveness.
Original languageEnglish
PublisherAalborg University
Number of pages128
ISBN (print)978-87-7152-049-1
StatePublished - 2014
ID: 208758337