ISAC商用化困局：AI加持通感一体，运营商变现路径待解

Integrated sensing and communications (ISAC) promises to turn mobile networks into radar-like sensors which detect drones and other objects, with AI turning raw sensing into decisions, but operators still need to find to way to monetise it. The need is clear, particularly for defence and public safety, but how operators will deploy and sell the capability is still unresolved, according to a panel at TIP Fyuz 2026 on ISAC and AI-native networks. The US Department of War (DoW) is putting money into early work through its FutureG programme, said senior technical advisor Scott Fox ( pictured, second from left ), among other reasons because it wants commercial technology which scales. “We design really exotic, very sophisticated equipment that’s really good, but it doesn’t scale, it’s hard to support,” he said. Operator-run ISAC would give it “a tiered defence that could tip and cue to higher, more exotic-type capabilities,” at a level of coverage the department could not afford to build itself. Last week, the National Spectrum Consortium and the DoW’s FutureG Office announced more than $50 million in awards to 11 teams under the RAPID programme, involving nearly 30 companies and universities, including AT&T and Ericsson. But in the department’s view, Fox said, it is “expected that mobile network operators need to find their own use cases”. “We don’t believe ISAC is possible without AI,” Fox said, adding AI also poses the greatest risk to it, pointing to underlying threats. Challenges Kalyan Sundhar ( pictured, fourth from left ), VP and GM of wireless strategy and industry partnerships at Keysight, said sensing “throws a wrench into a lot of things.” Engineers need to model the environment as well as the link, including reflections off buildings and the drones or robots being tracked, and then refine those models at scale. Shahram Mohajeri ( pictured, right ), senior principal member of technical staff at AT&T, described an ISAC demonstration with Ericsson in the US state of Texas several months ago, in which multistatic sensing located and tracked several drones. A system built for communication was doing a radar’s work, with AI at the site filtering clutter, but he noted “it takes a lot of resources”. Turning the technology into a new source of revenue “needs to be looked into very carefully”, he added. Steve Papa ( pictured, second from right ), CEO of Parallel Wireless, said multistatic sensing lowers the cost per node but makes timing critical, so his company is trying to build the first nanosecond-synchronised cluster of sites. Looking up for drones is a different problem from looking down, because cell antennas tilt towards the ground. There are also issues around waveforms which need to be worked out. Fox said OFDM goes blind in high-mobility, high-Doppler conditions, and the department is funding tests of alternatives with Cohere Technologies . Amitav Mukherjee ( pictured, third from left ), CEO of Tiami Networks, counselled restraint. “Use the energy that’s already there in the environment,” he said. In his view, OFDM-based sensing works as an early detection system, with other sensors brought in when more precision is needed. ISAC takes shape Mukherjee said tying ISAC to autonomous systems is the fastest route to revenue, because “you cannot wait for a human to step in and tell your delivery drone where to go.” His company’s platform, which he described as a multi-modal world foundation model which detects changes and predicts events, is aimed at fleets of delivery robots and drones, where every human intervention costs the operator money. In simulations, he said, it can cut the time or number of interventions by up to 30%. Keysight’s Sundhar made a similar point, saying micro-Doppler variations and trajectory data can feed models which predict what a target will do next. By Papa’s count, US operators have around 35 million analog-to-digital converters (ADCs) live in their radios which could give them spectrum awareness. “You’ve got all these millions of ADCs. You have hundreds of millions on the device side,” Mukherjee said. “These are additional ADCs that could be used for sensing as well.” Standards are also progressing. Mohajeri said ISAC is enabled in 5G-Advanced, with the northbound exposure API roughly 75% defined and a specification due by March. It’s also expected to be a key element of 3GPP 6G specifications, but Papa was more critical of the process: “We can’t trust the 3GPP to get this right,” he said. Operator game plan To offer sensing as a service, Mohajeri said an operator needs an exposure layer which authenticates requests, plus a new sensing function in the packet core which turns each request into instructions for the right radios. It also needs to decide whether the service is always on, since sensing takes capacity away from communications, and work out roaming and the sensing of people who are not its subscribers. As for monetising ISAC, Papa said he as