Mobile edge computing brings networking and computing closer to the device or network edge. This improves key performance metrics such as latency, throughput, and cost. In this talk, we present Spectronn’s cognitive mobile edge computing technology developed with PSCR funding. Specifically, we present the key lessons learned in developing and deploying artificial intelligence (AI) driven video analytics applications on mobile edge. Dynamic optimization of wireless networking for remote cloud access and local computing at the edge for low latency are discussed. Technology demonstrations are presented to support the technical approach. The technology developed from our research was deployed at the 2019 Boston Marathon to support the Brookline PD. This technology transition will provide resilient access to mission-critical applications (e.g., voice and video calls) and services (e.g., video streaming and storage) even when the wireless links are intermittent or totally unavailable.
Meet the Team
Principle Investigator: Rajarathnam Chandramouli
Resilient communications and computing (during emergencies and day-to-day operations) are key metrics for public safety applications and services. Resilience includes: (a) the ability to intelligently mitigate radio interference and backhaul congestion (e.g., end-to-end optimized cognitive radio networking); (b) the ability to cache and access data, deliver applications and host services even when the backhaul link to a core data center has failed or is severely degraded; and (c) smart computing to prolong the battery life of ﬁrst responder mobile devices, mobile computing gateways, and mobile and wireless sensors.
During an emergency, very often, the general public competes with public safety organizations (PSOs) for wireless network resources, mobile applications and services. This puts severe pressure and variability on latency and bandwidth available to PSOs that share public cellular networks, in times of need. The same is true when several PSOs share dedicated wireless networks for real-time and bandwidth hungry applications. Therefore, building cognitive wireless networks that sense mobile networks end-to-end for adapting communications and computing decisions to (stochastic) interference, congestion, coverage and capacity, is critical.
“Fog” brings “Cloud” closer to the end users, i.e., to the network edge for hyper local access and services, thereby building resilience against core infrastructure failures. Therefore, the two major goals of this project: (a) resilient fog heterogeneous wireless networking (HetNet) and (b) resilient fog heterogenous computing, are addressed by the following R&D tasks:
- the development of mathematical optimization models, algorithms and protocols for multi-radio access terminal (multi-RAT) and multipath (fog) HetNet and computing
- the development, modeling and analysis of an integrated, software deﬁned fog network-ing and computing architecture, with an innovative data plane–control plane separation called software deﬁned access (SDA)
- the development of cognitive HetNet router to demonstrate and ﬁeld a realistic working prototype of the proposed fog system by leveraging COTS hardware and open source software stack
- ﬁeld trial of the proposed technology with Brookline (MA) police department, our long time collaboration partner, leading to rapid technology transition to other PSOs
These tasks address Routing and Mobility Across Heterogeneous and Opportunistic Net-works and, Data Management, Access, and Consistency areas within the Resilient Systems theme. The proposed HetNet devices (router+computing device) along with a LiteEPC will be deployed in police cars at Brookline PD and tested. Resilience and robustness of the system under diﬀerent mobility conditions will be evaluated. Extreme wireless/core network conditions will be emulated in the ﬁeld. The successful demonstration, pilot trial and product delivery to the Brookline PD is expected to inform future U.S. (and global) spectrum policies for public safety communications.