DARPA (2017-2019)


Project Description

RAIDER: Reconfigurable and Application Independent Design for Radios (2017 - 2019)
  • Client:

    Project Sponsor- DARPA
Current SDRs allow selection of different waveforms, modulation schemes, front-end bandwidth, signal processing blocks, among others, and yet modifying these features through learning remains unexplored. Moreover, significant effort is spent in re-writing code as applications evolve. Finally, there is also no established representation semantics for information sharing among multiple devices and protocol layers within an SDR. RAIDER seeks to create a true ‘cognitive’ SDR platform that will enable resilient communication, increased RF environmental awareness, and intentionally protect or degrade reception capability at distant locations through three major research goals: Goal A- We will design a core set of re-usable software processing blocks that can be flexibly combined to meet high level objectives. Goal B- We will create a unified knowledge base to capture cross-layer device-specific sensed data, as well as feedback gathered from neighboring peer-devices spanning both signal-level (e.g., RSSI, phase/frequency offsets) and complex data formats (e.g., bit/packet error rate, interference level, battery level). Goal C- We will design a two-stage decision engine to initialize the software blocks- first by solving a multi-objective optimization problem, and then using machine learning for dynamic reconfiguration, using contextual knowledge as the network states get explored over time.

Fig. 1. Overview of RAIDER three high-level goals.

RAIDER will provide a deep theoretical understanding of how to create an application-independent SDR that can rapidly switch functionalities over time by maximally re-using high-level blocks. It will result in the first `cognitive SDR’ that goes beyond dynamic spectrum access, and reveals cost/benefit insights of brute force optimization versus machine learning-driven solutions to accomplish changing user objectives. The modular approach, open-source release of the code-base, and semantics for information exchange will bootstrap SDR development activities in a many intersecting disciplines.