Academicians at the Moscow Institute of Physics and Technology have created a prototype of an optoelectronic framework of unmanned vehicles for constructing terrain maps and subsequent movement along them.

The outcomes of the inquiry are promulgated in the Russian specialized publication Scientific Instrumentation. The new maturation differs from its analogues in that it can do without satellite navigation tools, and in some situations — without lidars. The work examines wheeled vehicles, but the framework is scalable and applicable on aircraft, as well as in water technology.

The equipment is installed on the roof of the vehicle and collects data about the surrounding environment. Information from its sensors and optical cameras is conveyed to the on-board computer framework, where it is analyzed, a three-dimensional terrain map is constructed and the car is localized. Then, grounded on the data obtained and the routes laid down, the framework plans further actions.

One of the important advantages of the new complex is the competency to do without lidars. The latter employ short laser pulses to gauge distances to objects. However, such devices are either expensive for mass use or have low accuracy.

Instead, in the proposed prototype, the developers created a framework of several multidirectional stereo modules. Each of them consists of several cameras combined into stereo pairs. These devices capture the image from different angles, subsequently constructing a three-dimensional model. The design and composition of the framework make it possible to regulate the space surrounding the drone. Also, applying the framework, it is possible to construct a map of the area, determine the location of the vehicle in it, analyze the information received and plot the further trajectory of the car.

Data processing is executed in real time and such controlled movement can be conducted without external navigation tools, explained Dmitry Gavrilov of MIPT. This sanctions the employment of unmanned vehicles in unfamiliar or unprepared terrain for tasks akin to agricultural work.

"We are making a framework that is capable to navigate solely by video image. This is beneficial in sites where there is no access to satellite navigation, for example, under a roof, where high—precision lidars or navigation receivers cannot be utilized due to their high cost," Dmitry Gavrilov said.

At the same time, the authors stressed that the matured solution is a modular design, which sanctions employing it on various types of equipment — from small robotic rovers to large trucks or combines.

It is also paramount that the framework sanctions constructing three-dimensional maps of various terrain types such as indoors or outdoors. In the future, it is possible to release versions adapted to work in conditions of limited visibility, equipped with thermal imagers, radars and other sensors. Experimental work on fusing data from different sensors has already been executed.

As robotics and autonomous systems spread to more domains like agriculture, disaster response and infrastructure monitoring, robust navigation devoid of reliance on external infrastructure akin to GPS will become more imperative. Modular solutions akin to this optoelectronic prototype that can fuse data from cameras, lidars and other on-board sensors exhibit substantial promise to address this necessity across robotic platforms.