The COMPUTEX Keynote series continued on Thursday with a presentation from Dr. Tzi-cker Chiueh, General Director of the Delta Research Center, that illustrated how Delta Electronics — already well-known for its power and thermal management solutions — is developing AI-powered strategies for a wide range of applications.
Delta are working to create innovative solutions in fields including wireless communication, industrial automation, building automation and smart mobility, Chiueh said, before presenting three recent case studies.
The first example was that of WiFi, which was originally designed for wireless communication but can now be repurposed as a sensor technology that accurately determines the position of WiFi devices. By calculating the time of flight and angle of arrival of a WiFi signal emitted from a device, its position can be triangulated.
However, the access point cannot determine whether these signals are arriving directly or after bouncing off a reflector: Delta use AI algorithms to solve this problem to within a margin of error of 0.35 meters. This method is so accurate that it can be used to detect a human’s breathing rate and heartbeat with 95% accuracy (within 5 meters) and 83% (within 1 meter) respectively.
This solution could have myriad applications, such as personal smartphone locationing, domestic intrusion detection, monitoring of sleep quality, and child presence detection in automobiles, Chiueh said.
A second case study concerned bridge inspections, which can now be automated with the aid of drones and AI. After creating a 3D model of the bridge segment to be inspected, an automated flight trajectory can be planned and executed even in the absence of GPS. The drone identifies inspection elements and slows down for precise imaging where anything suspicious is detected.
By automating the bridge inspection process, costs can be reduced by up to a half, while the total time can be reduced to one third of that required by a manual inspection, Chiueh said.
Finally, Chiueh demonstrated how AI is being used to stress test Automated Driver Assistance Systems (ADAS). In order to function safely and reliably in the real world, ADAS must first be rigorously tested in a virtual environment so that they are capable of responding to any scenario that they might encounter.
The AI systems developed by Delta are capable of generating a comprehensive set of test cases with which to cost-effectively stress test the ADAS. Next, they give feedback to help it improve performance, including by reporting whether it passes or fails the test, determining which part is responsible for failure, and generating a solution for next time.
