Unmanned flying vehicles (drones) are increasingly adopted for a variety of emerging applications ranging from rescue support to climate change monitoring. Drones can operate like mobile Internet-of-Things nodes, exchanging information and processing data onboard to minimize energy-consuming data transmission at long distances. In many such applications, the digital processing electronics on the drone may operate in extremely adverse scenarios, as for environmental conditions, energy procurement, inaccurate data sensing, along with heavy onboard computing workload. At the same time, cost limitation is essential to allow many civil applications that cannot usually afford high financial budgets, unlike military applications.
The proposed research addresses the design of specialized microprocessors and microprocessor systems, to support such a demanding technical context. The proposal leverage the merging of different orthogonal techniques for fulfilling the task, namely
- fault-tolerant hardware and software microarchitecture techniques, to allow commercial off-the-shelf technology to be employed in place of high-cost shielding materials and electronic devices;
- approximate computing techniques, to take advantage of the inherent inaccuracy of data sensing for saving energy;
- hardware acceleration techniques to allow the energy-efficient execution of heavy computational load.
The target hardware technology is FPGA, in the view of allowing in-flight re-configurability for micro-architecture tuning from the ground. A working prototype is going to be demonstrated at the conclusion of the activity, also in the context of larger research programs.
The innovation achieved by the proposed applied research activity relies in the unprecedented integration of specialized features in the new microprocessor subsystem, namely the approximate data support, the hardware acceleration, the fault tolerance. All of those characteristic will be tunable and adapted to the operating conditions during the mission of the drone, thanks to the configurability of the FPGA implementation.
Thus, the distinctive characteristic of the microprocessor cores will be the design-for-configurability approach, offering parameterized redundancy to the programmer. Not only the software application can be uploaded to the drone on-boad computer but also the hardware microarchitecture can be partially adapted from the ground to the specific effects experienced during the mission.