Learning from Data – The Search for Optimisation for Robot Networks
First author of a groundbreaking paper in the Proceedings of the IEEE, Dr. Hassan Jaleel’s work describes optimal solutions to complex problems linked with multirobot architectures.
Like swarms of sardines, swimming in a magnificently choreographed movement, robots in the future may need to form functional clusters that are adaptable to a multitude of requirements ranging from transport to mapping and surveys, exploration and communication.
One of the areas of interest for Dr. Jaleel’s study is task allocation and optimizing an underlying objective for multirobot systems. His recent study discovered a connection between coverage problems and task assignment. Like sardines in sea, multirobot systems may need to learn how to effectively perform decision-making using local data, to evade a pursuing target, for example. To deal with similar scenarios, Dr. Hassan Jaleel, along with his collaborator Dr. Jeff Shamma, took a jab at the multirobot problem using two approaches.
First, local decisions and global optimisation are seen from the framework of convex optimisation which is a well-established tool in locating most suitable solutions to complex problems. Second, the task of optimisation is linked with recurrent incidents of learning, framed in the language of games, where under prescribed rules and multiple players the best configuration is achieved by repetitive learning process between players.
The hope is that this work can serve large scale societal purpose and sets the stage for applying control and game theory to multiple scenarios. We congratulate Dr. Hassan Jaleel on having his paper published in this prestigious journal.
H. Jaleel and J. S. Shamma, "Distributed Optimisation for Robot Networks: From Real-Time Convex Optimisatiofn to Game-Theoretic Self-Organisation," in Proceedings of the IEEE, vol. 108, no. 11, pp. 1953-1967, Nov. 2020, doi: 10.1109/JPROC.2020.3028295.