Note: This content is accessible to all versions of every browser. However, this browser does not seem to support current Web standards, preventing the display of our site's design details.

  

Digital Fabrication

Most of the tasks during the construction process are of collaborative nature, as for example the manipulation of heavy or large objects. Therefore, it is necessary to design multi-agent robotic systems.

Cooperative manipulation

Within a multi-agent robotic system, each robot has a certain degree of information about its own and the cooperating robots' states and acts in order to achieve a common goal. Tasks of the involved autonomous robots may include sensing the environment, planning motion trajectories and ensuring safety of humans on the construction site. The control goal for the robotic system is to achieve good performance in cooperatively accomplishing the assigned task. Hereby, constraints due to hardware restrictions, environment and task accomplishment, as for example limits on interaction forces, have to be satisfied. Different group architectures for cooperative robots are possible:

Centralized control

The control inputs for each robot is computed in a central unit. An important drawback of this architecture is that the system is vulnerable to a single point of failure. Moreover, real-time control requires communication of each robot's system states to the central unit at a suitable frequency, as well as adequate computational power and memory capacity.

Decentralized control

The individual control laws are only based on information about each individual robot's states. Coupling interaction is taken into account as disturbance. This approach is rather conservative and task performance of the overall system is limited.

Distributed control

To overcome the drawbacks of the afore mentioned architectures, the following approach is possible: each robot has information about its own local states and communicates with its neighbors. Neighboring robots are defined by coupling states or coupling constraints between them. The individual local control laws are based on local (distributed) optimization, taking into account local information as well as coupling interactions to neighbors. This approach is most promising in terms of task performance, modularity of the system as well as robustness to failure. More information on cooperative control can be found here.

Uncertainties

Due to imprecise measuring and modeling, the robot's local information, as for example its kinematic parameters, configuration and dynamics are subject to uncertainties. Moreover, loads and contact forces due to interaction, as well as the positions of obstacles and of the collaborating robots are not precisely known. To face these uncertainties, different control strategies are possible:

Robust control

Robust control strategies are designed to avoid failure in the presence of measurement imprecisions or material tolerances. More information on robust control can be found here.

Adaptive control

Adaptive control strategies can be used. They are designed to dynamically correct the mathematical model, which does not exactly fit the real system due to inexact parameters or unmodeled dynamics.