Mobile Robots Localization and Map Building
Localization is the process of mobile robot pose estimation in real time relative to its environment which is represented by a map. The map of the environment is built using information gathered by robot perception sensors. The environments maps can be metric or topological or hybrid ones. In our research we use all three map types, where emphasis is given to their integration into a hybrid multi-resolution map in order to get the most compact robot environment description, with the possibility of novelties detection in the space and finding optimal strategies for incorporating these novelties into existing map. The main research problems in robot localization are: development of the algorithms for robot localization based on map features, distinction between persistent and non-persistent features and ensuring robot global localization in case of perception similarity of two or more subspaces.
In most cases mobile robots still require human assistance for exploration of first time visited environments, because they cannot self-localize without environment map, and consequently, cannot build the environment map without knowing their location. As the autonomous exploration of unknown environment is very important for commercial use of mobile robots, this problem is specially addressed in our research. Methods are investigated that ensure complete exploration of available space, with criterion of minimum exploration time, while suppressing the uncontrolled increase of the uncertainties in the robot position estimate and the environment map. Simultaneous localization and map building algorithms (SLAM) are investigated, with the goal to find an algorithm that ensures acceptable convergence rate of this dual process. The methods used for research of these problems include use of optimal stochastic estimators, decision theory and artificial intelligence.
Path Planning and Motion Control
The path planning algorithm generates optimal robot path. Path
planning for mobile robots and vehicles is mainly approached by two
concepts: reactive and model based approach. Reactive approach uses
only actual information from perception sensors, and is suitable for
dynamic environments, but cannot find optimal path in static
environments. On the contrary, model based approach is based on
previously built environment map and is suitable for optimal path
planning in static environments. We investigate hybrid algorithms
that use advantages of both approaches, with the goal to develop a
numerically acceptable algorithm for optimal path planning,
considering limitations caused by temporal changes in robot
environment (e.g. other moving robots, vehicles, humans and
animals). The algorithms that enable the tracking and the motion
prediction of moving obstacles in the robot environment are also
researched, which is very important for achieving optimal robot
motion in dynamic environments.
The motion control algorithms should ensure reliable tracking of the
path generated by the path planning module, considering kinematic
and dynamic constraints of the robot. As the drive mechanism of most
mobile robots has nonholonomic constraints, and also other moving
objects can obstruct planned robot path, it is necessary to apply
adaptive and/or robust control strategies. The optimal motion of
mobile robot through dynamic environment, in terms of smooth and
stable motion, and/or time or energy optimal motion is possible only
by the integration of motion planning and motion control algorithms.
The main problem of developing such an algorithm is high
computational complexity, which is caused by the mobility of the
robot. In order to overcome this problem, the development of the
algorithms that can be implemented with use of parallel processing
Control of Multi-Robot Systems in Intelligent Spaces
In numerous applications it could be more effective to use multiple simpler and cheaper mobile robots instead of one complex and expensive robot. There are many concepts in control of multi-robot systems, and we investigate the concept of multi-robot systems in so-called intelligent spaces.
Intelligent space is an environment with distributed sensors (e.g. cameras, microphones, sonars) and actuators (e.g. mobile robots, manipulators), with the purpose of providing various advanced services to the space users. The sensors are used for the detection and tracking of objects and persons in the space and for receiving orders from space users, and the actuators are used for delivering of physical services to the users (e.g. carrying or delivering loads). In terms of control and coordination of higher number of mobile robots, the intelligent space offers several advantages compared to the use of independent autonomous robots: a) it is not necessary to build an environment map, b) in every time instance it is possible to uniquely determine the position of the robot, c) even in dynamic environment, it is possible to find globally optimal path, because the sensor system has always perception of the whole environment, d) the whole intelligence is located in the environment, instead in the robots, so that robots share common resources and this results with lower system price.
In this topic, our research is mainly concerned with the robot identifications and global vision based position tracking algorithms, and also motion planning and motion control algorithms, with the goal of cooperative performing of more complex tasks.
Today's robots are mainly programmed using dedicated languages that require good operator routine,
and have very limited degree of programming flexibility. Extending mobile robot applications outside
the industrial domain require development of flexible interfaces for the human-robot interaction so it
would be possible for "common" users to use it in everyday life.
Our research focus is on the following aspects of interaction: a) teleoperation control of robots
and vehicles is oriented to realizing of visual feedback using virtual and augmented reality and force
feedback, so that the operator has the maximum feeling of presence in the remote location; b) interaction
via a database that is available to human on one side and to the robot on the other side - this is important
for mobile robots operating in automated shops, warehouses, hospitals etc.; and c) so-called intelligent
interaction via robot sensor system, where artificial intelligence techniques are applied to gradually
develop robot interaction capabilities during its cooperation with humans - this is important for the
applications where the robot helps a human being to perform various tasks (e.g. robots that help invalid
persons, driver assistance systems, etc).
Robotsoccer brings together the world's leading research labs in artificial intelligence (AI), robotics, and simulation. Research scientists compete in this sport as one way to test new technology, to explore the integration of artificial intelligence and robotics, and to exchange the latest developments in their fields.
To the casual observer, the game itself may seem like entertainment, yet it has proven itself as a valid platform for the evaluation of the performance of intelligent systems. Furthermore, the platform offers a chance for the evaluation of scientific results in fields such as systems control engineering, computer graphics and digital imaging, artificial intelligence, autonomous multi-agent systems and sensor technology. As such, Robot Soccer is more than just a game. It is a scientific tool that strives to find a solution by combining differing fields of research.
Our "ACT Croatia" team is participating in FIRA Mirosot League. For more details visit the team web page http://www.act-croatia.net'