Intel and Altera Expand Collaboration to Simplify Sensor Integration for Physical AI

Intel Corp. and Altera Corp. said today they are jointly developing flexible, real-time sensor bridge reference systems for physical and edge AI applications, including robotics. By combining Intel processing, connectivity and software with field-programmable gate arrays, or Altera FPGAs, the companies said they want to help manufacturers address the growing complexity of connecting, processing and acting on real-world data in increasingly sensor-rich environments.

“Physical AI systems must perceive, decide and act in real time, and the physical world requires precision, reliability and security,” said Farhad Shafai, head of business solutions and marketing at Altera. “Altera FPGAs provide the sensor connectivity, data processing and deterministic control that machines need to respond to the physical world, and remain programmable as sensors and AI models evolve. The combination with Intel processing and software is intended to provide OEMs with a flexible foundation to customize systems to their needs.”

Founded in 1983, Altera provides high-performance, programmable computing and connectivity systems for a broad range of advanced, AI-enabled technologies. The San Jose, California-based company said
is “the largest pure-play FPGA supplier with a full-spectrum portfolio serving applications ranging from data centers and communications networks to critical embedded systems.”

Intel acquired Altera for $16.7 billion in 2015 and sold 51% of the company to Silver Lake last year.

Sensor-rich edge environments pose challenges for AI

“Edge and physical AI systems may require data from different sensors and interfaces to be brought together while meeting timing, performance, power and cost requirements,” Intel and Altera said. “Some applications also require very high bandwidth or flexible form factors where input/output (I/O) is located far from the compute board. Sensor data must then be quickly streamed into memory for the host Intel processor to use within edge, robotics and other workloads.”

The companies said they plan to develop modular reference systems that combine Altera programmable hardware with a coherent foundation of Intel compute and software. Their goal is to give equipment OEMs more flexibility as applications, sensors and workloads evolve, while helping them move more efficiently from prototype to deployment.

“As sensors proliferate at the edge and across physical AI systems, our customers need an easier way to turn real-world data into timely action,” said Dan Rodriguez, corporate vice president and general manager of Intel’s Edge Computing Group. “Together with Altera, we aim to provide OEMs with a flexible foundation that simplifies integration and helps them move from prototype to market deployment more quickly, while meeting their performance, power and cost needs.”

Altera, Intel will apply real-time programmable processing

Intel and Altera have said they want to address more markets where real-time computing and programmable computing matter most. The companies intend to make the resulting reference solutions accessible to developers through Intel’s Open Edge Platform software development kit (SDK) and an Altera SDK.

• Robotics: The partners said they will combine Intel compute, connectivity and robotics software with Altera FPGAs to meet the increasingly complex sensing, timing and real-time processing requirements of intelligent machines.
• Healthcare: The companies are jointly developing a reference system aimed at the rigorous latency and real-time requirements of advanced healthcare equipment.
• Aerospace and defense: Intel and Altera are collaborating to support next-generation mission systems, with a focus on size, weight and power requirements for converged artificial intelligence and signal processing workloads.

Together, Intel and Altera said they will help OEMs move faster from prototypes by reducing integration complexity and meeting performance, power and cost requirements across a range of edge applications.