Research on architectures at the Department of Electronics and Telecommunication of Politecnico di Torino revolves along three different axes, namely extreme parallelism, unconventional computing, as well as telecommunication networks.

Along the first axis, we are investigating highly parallel heterogeneous architectures, based on a mixture of ASICs, FPGAs and GPUS. They are aimed at the analysis of images from camera and radar sources, at cryptography techniques that can counter the threat from quantum computing, at machine learning acceleration, and at efficient video and communication channel encoding and decoding. Design methodologies include a focus on using high-level synthesis to ease the re-use and porting of code developed for CPUs or GPUs on FPGAs or ASICs. Machine learning acceleration targets Deep Learning and exploits both digital and analogue techniques. We are also looking into smart embedded systems in an industrial IoT context.

Along the second axis we are studying memcomputing, approximate computing, analogue and mixed-signal design. Memcomputing imitates the organization of the human brain, to completely avoid the memory barrier and achieve extreme parallelism.  We are using approximate computing architectures as well as AMS design methods to optimize cost, performance and energy consumption for specific applications, including signal processing. A different form of non-conventional computing (at the human level) is to exploit crowdsourcing to solve problems. In this domain we are looking at how to partition jobs among workers and track dependencies about them, as well as how to ensure security, timely completion, and fair pricing.

Along the third axis, we are investigating various aspects of the 5G network architecture, of ultra-broadband optical access, and of network monitoring and optimization in terms of both performance and security. For 5G networks we are considering both network softwarization and slicing, to enhance flexibility and reuse of network equipment. We also focus on optimizing performance and cyber-security of computer networks, by collecting data in real time from the network, and extracting knowledge via big data machine learning algorithms. We also study how similar algorithms can be used to process the huge amounts of data from connected autonomous cars. Finally, we are exploring how optical fibers can be used both to reach every home and inside datacenters.