Illinois Tech professors Joohee Kim and Ken Choi have secured $1.6 million in funding from HL Mando for a project aimed at advancing control systems for autonomous vehicles. Over 4.5 years, they will work on developing artificial intelligence models and optimized computing hardware.
Kim's focus is on creating AI models that integrate data from cameras and LIDAR sensors to assess road conditions, identifying hazards like ice or potholes. She will also employ generative vision language models to generate synthetic data where photo data is insufficient.
"The objective is to predict the condition of the road and use this to improve the safety of autonomous cars," Kim explains.
The AI model development involves three stages, starting with a camera-based classification model, followed by a segmentation model for real-life application, and finally incorporating LIDAR sensor data for enhanced surface condition detection.
"It is quite difficult to identify potholes and bumps using only camera images, especially in bad weather conditions such as rain, storms, ice, or snow," says Kim.
Choi's contribution lies in designing computing hardware tailored for these AI models. He aims for an 83% reduction in power consumption compared to generalized hardware options.
"Big vendor companies such as NVIDIA, Intel, Microsoft, and Google provide some general solutions for hardware design, but these are far from optimal for this specific application," notes Choi.
The project involves collaboration with ten other research institutions and companies. It seeks to develop energy-efficient hardware using field-programmable gate arrays customized to specific needs.
"We want to optimize fully based on the algorithm," states Choi.