Overview / Usage

Intelligent autonomous agents and robots are designed with the intent to solve various tasks, such as imitating human behavior, keeping the human safe, acting in physically hard to reach domains (e.g. underwater on in space), and performing monotonous tasks. They might be obviously apparent, e.g. personal-use drones, industrial robotic arms, or military vehicles; or less so, with the popularization of internet of things (IoT). These agents share the same fundamental goal: autonomously plan (and execute) their actions. Yet, the increasing demand for these smart systems means they need to operate on cheap and accessible hardware.

In addition, robots which are set in the real world are required to account for its uncertainty in order to provide reliable and robust performance. Possible sources for uncertainty include dynamic environments in which unpredictable events might occur; noisy or limited sensing, such as a distorted output of a camera and an inaccurate GPS signal; and inaccurate delivery of actions. Yet, uncertainty measures are expensive to calculate. Overall, the computational cost of the decision making problem can turn exceptionally expensive, making it challenging for online systems, or when having a limited processing power.

Methodology / Approach

In our work we develop methods and algorithms to allow computationally efficient decision making under uncertainty, in order to make better decisions and increase safety. According to the problem definition, we wish to select the most beneficial action, according to some measure, e.g. gaining information about the position of the robot in an unknown location. Our methods are general enough to be used in the solution of numerous problems, including autonomous navigation, sensor placement and active sensing, robotic arm manipulation, and even more profound problems such as dialogue management. In several demonstrations, our algorithms showed a significant improvement in runtime.

For an up-to-date list of related publications, visit:
www.khen.io or
www.researchgate.net/profile/Khen_Elimelech

Technologies Used

Technologies Used

Repository

http://www.khen.io