In Haoxin Wang’s eyes, once-futuristic visions of self-driving vehicles are closer to reality than many people think. His research is focused on trying to help revolutionize the design of those autonomous vehicles.
Advancements in self-driving technology in recent years have ranged from adaptive cruise control and braking to efforts to create fully autonomous vehicles.
Wang, an assistant professor in the Department of Computer Science at Georgia State University, is addressing the need for advanced vehicle computing power to make self-driving technology more accessible and efficient.
The key, Wang says, may lie in edge computing and sustainable artificial intelligence (AI) to improve the way smart vehicles operate.
“We want to make sure that every vehicle has sufficient computation power when they’re running AI applications like self-driving algorithms or image processing,” said Wang, who refers to the work as “protocol and algorithm design.”
“In the future,” he said, “our vision is that all vehicles can have fully autonomous driving functions.”
That would mean a more equal and fair computing environment, making it possible for vehicles to have the most advanced technology regardless of price.
Through their research, Wang and Jiang Xie, of the University of North Carolina at Charlotte, could change the way intelligent vehicles function through automotive edge computing.
Edge computing refers to a system of servers and wireless technology that works like a remote computer for a vehicle. This approach allows the computational work to be offloaded from the car to external resources.
Recent advances in car technology rely on downlink connections (transmission from a computer server to a vehicle) or uplink connections (transmission of information from a vehicle to a server). Downlink connections support features such as streaming entertainment. Meanwhile, features that require cars to take in information from the environment, such as self-driving technology, are supported by uplink connections.
In some cases, both downlink and uplink connections are used. For example, navigation technology may use uplink technology to transmit location data from the car to the remote server, and then use downlink technology to send the updated route information back to the car.
However, current vehicles are limited in computing power. As a result, manufacturers must choose how to allocate resources that support each of these functions.
Traditional network resources are considered asymmetrical. That is, they allocate more resources to downlink applications than uplink applications. In the future, this could become an issue, as uplink applications are vital for self-driving technologies.
This is where Wang sees an opportunity for improvement.
Using an algorithm to enhance the performance of existing edge computing technology, his research proposes a new model of resource allocation that includes external edge servers.
“The vehicle will be like the data collector,” Wang said. “It collects data from the surrounding environment and transmits the data to the edge server for processing. After it finishes processing, the edge server will return the results back to the car.”
Offloading the computational process to an external server would make the vehicle’s hardware limitations less restrictive. This technology could provide more equality across vehicles, regardless of their built-in computing power.
In addition to making these forms of technology more accessible, Wang wants to make them better for the environment. He speaks in terms of a sustainable artificial intelligence ecosystem.
“Most researchers today care most about the performance of AI, what the AI can provide,” Wang said. “What we’re caring about right now is the sustainability of AI.”
AI applications require a large amount of data and power, resulting in substantial energy consumption and carbon emissions. Wang is striving to make AI more environmentally friendly so that as it becomes more prevalent in the future — perhaps with the help of his research — it can become more sustainable.
“This kind of environment-friendly AI is very important for the future of AI. We need to make it environment-friendly and sustainable enough to support our community,” Wang said.