July 21, 2022
Technion researchers solving essential problem in AI field – study
A robot equipped with artificial intelligence is seen at the AI Xperience Center at the VUB (Vrije Universiteit Brussel) in Brussels, Belgium February 19, 2020. (photo credit: REUTERS/YVES HERMAN)

The study is related to a problem in AI and robotics called autonomous decision-making under uncertainty.

Researchers at the Technion – Israel Institute of Technology found a way to simplify decision-making and problem-solving under uncertainty in a way that reduces the amount of information computers need to analyze.

A new peer-reviewed study published in the International Journal of Robotics Research, led by Prof. Vadim Indelman, who heads the Autonomous Navigation and Perception Lab (ANPL) at Technion’s Faculty of Aerospace Engineering, and Khen Elimelech, shows the feasibility of reducing the amount of time for computers to process information without compromising the success of completing a function.

“We demonstrate that we can significantly reduce computation time, without harming the successful execution of the task,” the researchers said. “We also demonstrate that computation efforts can be reduced even further if we accept a certain loss in performance loss that our approach can evaluate online. In an age of self-driving cars and other robots, this is an approach likely to enable autonomous online decision making in challenging scenarios, reduce response times, and achieve considerable savings in the cost of hardware and other resources.”

The study is related to a problem in AI and robotics called autonomous decision-making under uncertainty, which concerns the capability of AIs to complete tasks reliably and autonomously over time in an unpredictable environment.

Technion noted that autonomous agents often do not have access to the variables related to a particular problem and instead function based on a “belief” based on probability models and measurements.

Illustrative image of AI. (credit: PIXABAY)

Belief space planning

A major area of research in the new study was computationally efficient decision-making under these conditions, called belief space planning. In order to solve this problem, an AI must weigh the costs and benefits of a potential action, which requires the researchers to predict how the “belief” will change over time.

The findings may help researchers solve decision-making problems using simplification and show that there are ways to save considerable amounts of computation time without a loss of accuracy.