NXP has Freescale's S32V234 processor, which integrates the Cognivue’s computer-vision IP core, while Qualcomm has its Zeroth deep learning platform.
NXP CEO Richard Clemmer’s comment about his company’s Machine Learning lag took the automotive market somewhat by surprise.
At issue is not just machine learning. How about artificial intelligence and computer vision? How far is NXP into the R&D and product development of such areas? Demler told EE Times, “You can put all the AI/machine-learning/computer-vision topics in the same basket.”
He said NXP has the former Freescale S32V234 processor, which integrates the Cognivue’s computer-vision IP core. [Freecale acquired Cognivue before NXP bought Freescale].
Meanwhile, Qualcomm has “their Zeroth deep-learning platform, which combines their CPU, GPU, and DSP cores,” Demler noted.
“Cognivue and Freescale were already established in automotive, so the question is whether any customers have adopted the S32V234,” he dded.
Similarly citing Qualcomm’s Zeroth platform and its machine learning research projects, Krewell said, “The company was ahead of NXP in recognising that machine learning was going to an essential part of autonomous driving and smarter devices.”
Krewell, however, cautioned: “My concern is that this deal [with NXP] will slow Qualcomm's work in this area as it places more resources into the NXP merger.”
Asked about Qualcomm’s vision technology, Webber said, “In automotive, nothing public has been shown.” He added, “However, in 2015 at Google’s developer conference Qualcomm came out with its own reference designs for computer vision based depth sensing in phones equipped with Qualcomm’s Snapdragon mobile processors. AR & VR are development areas for Qualcomm.”
Meanwhile, Demler remains more optimistic. The trend in embedded computer vision is to combine general-purpose processor cores with special-purpose vision accelerators, Demler said. “So, the combined companies now have the opportunity to use the best of each other’s IP.”
Radio war over V2X
The battle over the radio choice for V2X is represented by two camps—those who push IEEE 802.11p/DSRC and those who advocate LTE.
The battle has been brewing in Europe, but also in China, one of the world’s most important emerging automotive markets.
Along with telecom technology giants such as Nokia, Ericsson and Huawei, Qualcomm has been a leading voice for LTE-based V2X.
Demler, however, is sceptical. “Qualcomm has a ‘vision’ of LTE-V2X, but they’re going up against already established standards that trace back to the FCC’s allocation of 5.9GHz ITS spectrum in 1999.” He noted, “Qualcomm is the leader in LTE modems, so of course they prefer to use it anywhere they can,” he added.
Demler believes, “Although there’s a place for LTE in smart transportation systems, the whole industry, standards groups, and government agencies aren’t going to bend to Qualcomm’s vision.”
Strategy Analytics’ associate director, Roger Lanctot, shares a different take.
He sees the V2X radio battle as “ultimately a moot point” because of the complex web of partnerships it has already entangled.
For example, “Qualcomm is knotted up with Cisco over the question of channel allocation for DSRC spectrum,” Lanctot noted.
“Qualcomm already owns DSRC supplier Atheros and may now own NXP which is a big proponent of current DSRC specs/allocations,” he added.
NXP’s position pursuing DSRC-based V2X has been reinforced via its NXP’s investment in Cohda Wireless. Cohda makes IEEE 802.11p/DSRC-based roadside units and on-board units.
But then, “Cohda is now testing on 5G, which Qualcomm supports,” said Lanctot.
In his opinion, “Whether [President Barack] Obama moves the V2X mandate forward or not, V2X technology will progress.” That can be done either over dedicated devices [based on DSRC] or over 5G and advanced forms of LTE, he explained.
In the end, he concluded, there’s no real impact—except from a lobbying perspective.