Coordinated Task Allocation in Multiple Autonomous Underwater Vehicles under Information Constraints
The main works of the joint PhD study will be focused on challenging aspects as follows:
(1) Study the control methodologies for the multi-AUV system as network topologies play an important role in the coordination of robots. An AUV group, in which every AUV could get the information of the other AUVs either directly or indirectly through the communication topology, can choose to or not to form a communication coalition which can be controlled in a central unit (two AUVs can communicate with each other directly if the distance between them is within the communication distance radius). In that way, the situation awareness of each robot in one coalition can reach a consensus through the connected communication topology, thus leading to the consensus task allocation locally in each coalition . It also could improve the computational efficiency of the whole network. However, many other questions need to be resolved regarding whether the bigger coalition size would contribute to the better performance and the related network stability, as well as what the best principle is of choosing the central control AUV in each coalition.
(2) Design adaptive strategies to direct the communication topology evolution of the multi-AUV system and study eigenvalues of the Laplacian matrix of the connected multi-AUV network in order to make the task allocation consensus converges quickly. Firstly, because of the movement of AUVs, the communication topology of each coalition changes all the time, thus leading to dynamic abruption of the old coalitions and formation of new coalitions. As a result, adaptive dynamic protocols need to be designed to obtain consensus. Secondly, the research shows that the eigenvalues of the Laplacian matrix of the network are important for the convergence speed of the multi-robot consensus problems  . Issues to be dealt with are how to decide the movement of the AUVs in order to keep the old coalitions or form new formations to generate proper eigenvalues of the network Laplacian matrix. Once a coalition being formed, another important question is how the task allocation should be triggered (i.e., event-triggered or time-triggered). Whatever manner is chosen, the updating frequency of the task allocation should be studied in case it is too high to maintain the stability of the robot network or too low to prevent resource waste.
(3) Propose task allocation algorithms that take the marine robots’ path planning into consideration. Firstly, as described above, AUV path planning problem couples closely with the task allocation taking ocean currents into account. What’s more, in many cases the marine robots are coupled together through the tasks they are required to accomplish, but are dynamically decoupled by assuming that the motion of one robot does not directly affect the others. Thus, when multiple robots move within a common space, they should attempt to avoid collisions . Fundamentally, this may be viewed as a problem of resource conflict, which could be resolved by introducing, e.g., traffic rules, priorities, or communication architectures. In conclusion, path planning must be performed while taking into consideration of other robots and the global environment. This multi-AUV path planning, couples closely with the task allocation, is an intrinsically geometric problem in configuration space-time.
|Last modified:||06 September 2019 3.35 p.m.|