In this research work, we focus on the development of novel behavior based robot control methodologies. In a behavior based context, a complex control problem is subdivided into a number of more simple problems, called behaviors. The question arising in this situation is how to fuse all the different behaviors to form one consistent global robot behavior. This is called the problem of behavior fusion or action selection. In this research work, we investigate novel approaches to solve the action selection problem in a generic way. These new methodologies must seek to eliminate two of the most prevalent shortcomings of existing action selection techniques: they must require a minimum of parameter tuning and they must make it possible to incorporate knowledge from a human expert in the control loop. Indeed, robots are more and more entering our everyday lives, meaning that the interactions between humans and robots must be taken into account in the development of the robot control architecture.
An important aspect of this research work is the experimental validation of the developed theorems on simulated and real robotic systems. To this extend, we set up a simulation framework to test and validate the proposed theorems and compare their performance with respect to existing approaches. We also apply the developed action selection methodologies to real robotic systems operating in outdoor environments. Working in completely unstructured outdoor environments, where external disturbances will always complicate the robot control problem, allows us to assess the performance of the developed approaches and their capability of dealing with large uncertainties in complex and large environments.
Behavior based control approaches have already proven successful in a number of robotic applications. Mostly, these applications consider some mobile robot wandering in an inside environment. However, as more and more robots become operational, the coordination between all these robots also becomes an issue. Therefore we also investigate in this research work the possibility to extend the application field of behavior based robot control methodologies to the field of multi-robot coordination. This means that we also develop behavior based control methodologies to make multiple robotic systems operate together as a homogeneous team. Application scenarios for multi-robot collaboration tasks are defined and implemented in simulation as well as with real robots.
The proposed behavior based control methodology will also be extended to the case of heterogeneous robot teams, including a mix of air and ground units. Mixing the capabilities of UAVs (Unmanned Air Vehicles) and UGVs (Unmanned Ground Vehicles) offers tremendous possibilities due to the different capabilities of both types of vehicles. On the other hand, this kind of hybrid robot control application complicates the robot control methodology even further, due to the very different nature of both kinds of platforms. In this research work, we will seek to extend the behavior based robot control methodology, such that it provides an integrated framework for the control of air and ground robotic units.
Video Results (check also our YouTube channel):
Robots used for this research subject: