Thousands of computer nodes running simultaneous traffic simulations in a cloud computing environment can predict how autonomous vehicles (AVs) and unmanned aerial vehicles (UAVs) should react to challenging situations, but cloud connectivity isn’t guaranteed.
A car in a wireless dead zone or a UAV operating in an electrical storm may be cut off from computing resources. As a solution, researchers at Washington State University (WSU) have created a software framework to bring some of that cloud-based artificial intelligence (AI) to smaller devices.
They presented their most recent work at the 2020 Design Automation Conference and the 2020 International Conference on Computer Aided Design, demonstrating a framework for more efficient use of AI algorithms on mobile platforms and other portable devices.
“The goal is to push intelligence to mobile platforms that are resource-constrained in terms of power, computation, and memory,” says Jana Doppa, George and Joan Berry associate professor in WSU School of Electrical Engineering and Computer Science. “This has a huge number of applications ranging from mobile health, augmented and virtual reality, self-driving cars, and digital agriculture, to image and video processing mobile applications.”
Voice-recognition software, mobile health, robotics, and Internet of Things (IoT) devices use AI, and AVs remain somewhere on the not-too-distant horizon. Keeping that decision-making on the cloud doesn’t work, Doppa says.
The cloud isn’t fast enough – a self-driving car deciding to turn right while looking both ways would require information to go from the car to the cloud and then back to the car.
“The time required to make decisions might not meet real-time requirements,” says Partha Pande, Boeing Centennial Chair professor in WSU’s School of EECS. He adds that many rural areas also don’t have enough connectivity for AI communications back and forth through the cloud.
However, running sophisticated computer algorithms on portable devices is also problematic. Phone computing memory is small, and a lot of decision-making will quickly drain battery power.
Doppa’s group came up with a framework that runs complex neural network-based algorithms locally using less power and computation; the approach prioritizes problem solving. As in human decision-making, in which problems vary in complexity and require more or less brain power, the computer framework spends a lot of energy on only the complex parts of problems while using less resources for the easy ones.
“We are improving performance and saving a lot of energy,” Doppa says.
For example, in a digital agriculture application, more efficient software and hardware embedded on a UAV could efficiently make decisions about crop spraying with less computational and energy requirements.
Washington State University https://wsu.edu