Cell Phone Towers: Monitoring Tenant Activity On Site.
Artificial intelligence at the edge
The increasing data usage by consumers and devices (e.g. IoT devices, self- driving cars and smart cities etc.) has led to a competitive telecom landscape across regions, as a result, network carriers are opting for tower sharing in-order to expand their network coverage in developing markets and congested urban centers where tower site acquisition is difficult. This results in high-levels of site activities, presenting tower companies with the challenge of monitoring structure loading and equipment installation in order to prevent revenue leakage and impairments.
“The telecom towers market was valued at 4.82 million towers. In 2019 and is expected to reach 6.29 million towers by 2025, at a CAGR of 4.56% over the forecasted period of 2020-2025.”
The exponential growth in the number of tower structures presents tower companies with the challenge of monitoring these structures. Even more so for the tower companies which are optimising the tower sharing model; having multiple network carriers leasing antennas space on your cell tower. Tower overloading and illegal installations become an imminent problem for these tower companies, as some network carriers capitalise on the prolonged scheduled site audits which sometimes occur annually. Service Level Agreements in some cases would require the tower company to confirm network carrier installations before billing, this would result in high-travel costs in ad hoc site visits and revenue loss due to identifying the installations late.
American Tower Corporation (ATC), a leading independent tower company, with a global tower portfolio of more than 171,000 sites in over 16 countries and ranked 410th in the fortune 500 companies. ATC South Africa was the test ground for a case study for a remote tower asset monitoring system underpinned by artificial intelligence (AI) at the edge. The objective of the case study was to ascertain the feasibility of an AI system that can conduct visual inspection of the tower structure and exception reporting at the edge. Due to the scale of tower sites, bandwidth optimisation, automation and intelligent data management was a key driving factor to the selected system architecture. The scope of this case study consisted of the development of proprietary narrow AI software such as computer vision, self-powered cameras and edge servers.
Our findings were that our AI system was able to obtain an 95% accuracy in identifying anomalies and exception reporting. In addition our proprietary hardware devices, at a technology readiness level of 6, achieved key milestones to affirm the feasibility of the AI system.