The system consists of a quadrocopter, a landing platform situated on the roof of the car and a specially prepared software running on a notebook and the quadrocopter’s board computer.
|Quadrocopter starting from the platform situated on the roof of the car.||Our dedicated software supports multiple input devices to control the drone|
The platform is a place where the quadrocopter can be securely held while driving. The notebook is required in order to easily control the behavior of the drone and the platform.
Computer Science problems
Laser scanners are heavy and expensive, therefore computer vision should be used to detect obstacles and map the environment. The most intense, difficult and time-consuming area of research connected with this project is the development of visual environment recognition algorithms. Detecting obstacles is a fundamental aim of environment recognition. There are several obstacle detection methods but all of them are at the moment unreliable and allow only for slow maneuvers between obstacles.
Moreover, for efficient route selection, it is essential for mobile robots such as drones to map the environment. For the preparation of maps of the environments 3D SLAM algorithms are used. 3D SLAM algorithms generate an overwhelming amount of cloud point data and require enormous computational power to analyse it. A great solution to this problem would be developing an objective SLAM algorithm similar to SLAM++ . A member of our team tries to solve this problem in his PhD.
There are some traditional mechanical difficulties – high weight and cost, big size, low battery capacity. In 2030 the batteries will be more capacious and lighter. We are currently working on miniaturisation of the UAV and replacement of the LIDAR obstacle detection method with stereovision.
The concept of the future single rotor, coaxial urban UAV is illustrated in the following image. The single rotor UAVs are harder to steer and more difficult to build, however they are lighter and offer longer flights and better dynamic.
The possible uses of UAVs for things such as parking place detection and highlighting, traffic jam recognition and alternative route finding, overtaking support, low visibility places preview or emergency vehicle detection are only some of the many opportunities for using drones in cooperation with cars. The variety of possibilities provided by UAVs will allow the improvement in the functionality of the system by developing and adding new, interesting and practical applications.
The total cost of the prototype was around 1500 $. Approximately 90 % of the cost was the electronics, so there is a very wide area for cost reduction. All electronic components were bought separately, and they are protected by their own housing and are connected with relatively long and heavy connectors, therefore there is a great possibility for the weight reduction. All mechanical components were connected with steel nuts and bolts, thick aluminum brackets were used to link the solid aluminum profiles, so there is a great margin for reducing the weight and assembling time.
Our research team has been developing this project since May 2014. We are constantly improving our idea and prototype. We have a lot of ideas which are not yet tested and implemented. Based on the progress that we have made this year, we are able to say that our project is feasible. Our team manages a code repository, where the source of the software used in this project is hosted, constantly updated and available for download.
The most intense and time-consuming area of research connected with this project is the computer vision and artificial intelligence programming. Fortunately our research team isn’t struggling alone with these problems, there are some big, open source projects like Google Tango aiming to overcome similar difficulties.