SUMMARY
Researchers led by Tsang-Kai Chang and Chaojie Feng from the Department at UCLA have created a scalable and real-time SLAM algorithm for autonomous robots.
BACKGROUND
A mobile robot relies on the simultaneous localization and mapping (SLAM) algorithm to create a map in an unknown environment while also localizing itself in it. As robots become more integrated into everyday life, SLAM will play a key role in making these robots autonomous whether they do tasks primarily indoors, like a Roomba, or outdoors in areas with weak GPS signal, like self-driving cars. The SLAM algorithm models two key components, the spatial state of the robot and the positions of landmarks. The way these parameters are modeled separates the two current state of the art SLAM algorithms. The filtering-based approaches can estimate the two parameters in real-time. However, complexity of this algorithm increases dramatically with the number of landmarks and thus cannot scale to more sophisticated maps. The optimization-based approaches updates these two parameters offline and thus makes harder to process incoming information during movement, nor can it create an explicit map directly which is needed for other autonomous processes like path planning. In short, there is no SLAM algorithm that has low complexity with respect to number of landmarks, online feature extraction, and can build explicit maps.
INNOVATION
Researchers led by Tsang-Kai Chang and Chaojie Feng from the Department of Electrical and Computer Engineering at UCLA have created a scalable and real-time SLAM algorithm for autonomous robots. Their invention, named the OEM SLAM, combines the benefits of both filter-based and optimization-based approaches to create a robust SLAM algorithm. The OEM SLAM has many benefits over its predecessors which include online computation that allows it to work in real-time as the robot is moving, low complexity that allows it to easily scale for larger and more complex maps with more landmarks, and the ability to create an explicit map which allows the robot to perform higher level tasks like path planning. In a pilot study the OEM SLAM algorithm shows the ability to estimate landmarks accurately even when the landmarks are observed intermittently.
POTENTIAL APPLICATIONS
- Indoor Robots
- Augmented Reality Systems
- Self-driving Cars
ADVANTAGES
- Less computation cost
- Manageable scalability
- Real-time
- Ability to build explicit maps
STATUS OF DEVELOPMENT
Pilot study the OEM SLAM algorithm shows the ability to estimate landmarks accurately even when the landmarks are observed intermittently.