| Unmanned systems are becoming increasingly popular for dangerous, dirty, and dull tasks. Unmanned aerial vehicles (UAVs) allow operators to maintain a safe distance from the workspace while maintaining an aerial presence. Multi-agent systems not only speed up large tasks but have the potential to increase robustness to single-agent failures, ensuring mission success despite degraded equipment. This thesis studies the dynamic travelling repairman problem (DTRP) where a team of m vehicles is tasked with servicing a continuously increasing set of service requests over time.;Four desired properties are presented for unmanned multi-agent systems: reliance on local communications, reliance on local sensing, ability to operate through disrupted communications, and robustness to single-agent failures. The literature is surveyed finding no policies for the DTRP which satisfies all four properties. A novel collaboration framework is presented with practical constraints on the information transferred between collaborating agents. Two control policies are designed from this framework satisfying the four desired properties for unmanned systems. A bound for the asymptotic value of the expected wait time is derived and validated with simulation trials. The collaboration framework provides a platform for multi-agents systems with practical considerations for hardware implementation.;Keywords: Unmanned vehicles, cooperative control, autonomous systems, distributed systems. |
