Autodriver (Control Systems for Mobile Autonomous Robots)
Tech Area / Field
- SAT-EXP/Extraterrestrial Exploration/Space, Aircraft and Surface Transportation
- SAT-SUF/Surface Transportation/Space, Aircraft and Surface Transportation
- INF-SIG/Sensors and Signal Processing/Information and Communications
- ENV-MIN/Monitoring and Instrumentation/Environment
3 Approved without Funding
State Institute of Physics and Technology, Russia, Moscow
- CAP GEMINI, France, Meylan
Project summaryThe importance of the task of the development of autonomous mobile robots is determined by the necessity to work in hazardous environments that are fatal о dangerous for men (radiation, chemical contamination, high temperature and pressure etc.) and when teleoperation is impossible (in screened rooms, in mines, under radic noise) or when time of radio signal traveling is large (exploration of planets of the Solai system).
Main objective of the project is the development of algorithms and software modules for processing TV and 3D laser images with the aim to determine admissible speed and direction of motion.
The approach to the autonomous driving problem is based on a compute technology with the use of methods of cybernetics, synergetics, artificial intelligence computer vision, computer simulation, system theory, decision theory, pattern recognition, fuzzy sets, and fuzzy logic.
The salient features of the project that distinguish it from the wide range of similar works consist first of all in the complex approach to the problems of perception planning and control, and then in the fact that the methods that lie at the base of the project accumulate the long-standing (over 20 years) experience of fundamental applied, and experimental research. The prototypes of the algorithms to be developed were tested, selected and debugged, passed evaluation tests on real vehicles and proved their efficiency under real conditions.
Our know-how is a reliable base for the development of new algorithms and software to vehicles of various size and function, having various propulsion devices (wheeled caterpillar, wheeled-walking, etc.), and to various speeds and conditions (planets of the Solar system and the Moon, cross-country, roads, including country roads, and indoors).
As a result of carrying out the project, it is supposed to develop algorithms and software modules for:
• analysis of unprepared terrain scenes by 3D laser images;
• identification of dangerous areas (obstacles) depending on dynamic and crosscountry capabilities of the vehicle;
• working out the trajectory for negotiating obstacles detected on the unprepared terrain (at the tactical level of control);
• determining the admissible speed and motion control along the trajectory that was specified as a result of image sequence analysis; the control should take into account the dynamic characteristics of the vehicle;
• analysis of sequences of mono-TV images of road scenes; the algorithms should be resistant to noise (shadows, puddles, patches, blurred boundaries of the roadway);
• planning of speed and heading control for road following.
We expect that the ability of a surface to carry traffic will be determined by the system at the distance of up to 25 meters from the vehicle. Obstacles that tower over the terrain will be identified at the distance of up to 45 meters. The decision time for cross terrain is expected to be no more than 0.8 s. These characteristics of the system make it possible to achieve the speed of up to 10 m/s (36 km/h).
The time of a TV-image processing is expected to be about 0.25 s. The time required for planning the control actions to follow the road does not exceed 0.05 s. Thus, the overall time of the decision making will not exceed 0.3 s.
The modules specified generate a core of the middle (tactical) level of the motion control system. Not neglecting the importance of the upper (strategic) level of the control system that is based on a cartographic database and of the lower (executive) level, it should be noted that this is the tactic level that mainly determines the capabilities of the autodriver as a whole. This is in the development of this level algorithms that the researcher faces radically new scientific problems. These modules are supposed to be integrated into control systems of autonomous robots designed by the Foreign Customer.
Collaborators from the EC, the USA, and Japan that are planning or already conducting the development of autonomous vehicles can substantially accelerate the fulfillment of their projects (by several years) and gain advantage that (taking into account the high level of salaries in these countries) can amount to millions or even tens millions of USA dollars if the scale of the works is large.
Role for Foreign Collaborators
The authors of the project hold to the idea that the effectiveness of the methods and algorithms for autonomous driving can be thoroughly tested only within a real fully autonomous vehicle control system. Since presently it is impossible to perform such a test in Russia, the role of the Western collaborator seems to bae decisive in what concerns this issue.