We will be guided by case studies and specific design requirements for fleets of multi-autonomous air vehicles (referred to as UAVs or UC(Combat)AVs). While this may seem specific, the problems that we will encounter are in a certain sense generic to the control of several kinds of multi-agent systems (some civilian systems that we have had past experience with include Automated Highway Systems and Air Traffic Management Systems). The topics of the proposal are basic research in this new paradigm of distributed, software-based control for autonomous, uninhabited aerial vehicles. In this proposal, we highlight the following enabling basic research thrusts:
A second application area that may be pursued in the second phase of the proposed effort is control of gas turbine aeroengines. These machines consist of several interconnected aerothermo dynamic subsystems and engine controllers must creatively use very limited actuation power, be capable of dealing with engine-to-engine variations, sudden configuration changes (eg, due to faults) and degradation over the time between overhauls. While computers have been used for full authority engine control for many years, they have not yet capitalized on the rapid growth in capabilities in the computer industry at large. These systems are an excellent target for new software-enabled control designs since the basic hardware requirements are currently in place, and hence there is the potential of near-term impact of the tools to be developed under this effort.
In addition in our single agent research, we will primarily work on software tools for high-level planning and control traditionally not addressed in the domain of conventional control. High-level control and planning includes tasks such as navigation, mission planning, reconfiguration under failure modes (in the control or action functions) and sensor fusion, target acquisition, etc. (in the sensing or perception functions).
A prime deficiency of work in the area of software for complex control systems has been the lack of a team approach involving experts from specification, modeling, simulation, and experts in control hierarchies and multi-modal multi-sensor control systems to experts in software engineering and formal methods for the synthesis and verification of control software. In this project, we bring together a team with expertise in modeling and simulation (Lee), computer-aided verification and formal methods in real-time systems (Henzinger), control design for hybrid and multi-agent systems (Sastry), and software systems (Aiken) to develop methods and tools for the synthesis and analysis of control software for complex systems. Our team has experts in each of these areas, with a past record of working together, who will bring to bear their skills in creating software design tools for control systems where there is a combination of discrete decision making and planning under multi-modal multi-sensor architectures, with various levels of centralization.