Development of blockchain oriented distributed task sharing system architecture for multi robot systems

In multi-robot applications; operational processes such as robot management, task assignment to robots, task tracking, communication and coordination between robots, data storage are carried out through a central server system and the most critical decisions are made by this server without leaving them up to robots. However, centralized systems have several disadvantages such as the necessity of server maintenance, routing all communication through the central unit, making decisions from a center, the risk of a single point of failure, possible communication problems (especially major communication problems encountered in multi-robot systems in open terrain) affecting the functioning of the whole system, and increasing bandwidth requirements as the number of robots increases. In this thesis, a blockchain-powered distributed and decentralized system for inter-robot communication is developed and a task sharing application is implemented between robots working in this system in order to overcome these disadvantages. Thus, a dynamic and distributed architecture is introduced and implemented in which robots can communicate directly over a blockchain network without using a centralised server. Hyperledger Fabric (HLF) developed by IBM is used as the blockchain platform due to its advantages. Each robot in the system is defined as a separate physical peer node in the HLF network. Making task sharing, which can be easily done in centralized systems, to achieve an optimum level of "joint decision" is a major problem in distributed and decentralized systems. With the architecture developed in this thesis, robots use a "cost function" to select the most cost-effective tasks for themselves, while at the same time keeping the communication load in the network at an optimum level. After peer robots in the HLF network reach consenus on the task allocation, task data is transmitted to the ROS running on the robots via a socket connection. The task allocation process in the network is structured in such a way that only robots that are not "busy" at the time are included, thus preventing "busy" robots from affecting the process. Experiments conducted with real robots and in simulation show that the proposed system performs optimal task allocation for all conditions. It is envisaged that this thesis work can inspire the creation of a strong alternative to the centralized approach by introducing innovative contributions to the literature. The developed system is adaptable to all autonomous systems and has high potential in both industrial and military applications.