Vision Based SLAM
Simultaneous Localization and Mapping
A critical element for the operation of an autonomous system is the ability to navigate from one point to another. This stereo-vision technology is the core of Vision Robotics technology. The flexible system can map indoors and outdoors and can work with other sensors such as LIDAR to provide a robust mobile platform. The mapping technology itself may be used on any platform to deliver payloads and perform a wide variety of tasks.
The navigation uses a 2-D map collapsed from the 3-D grid using the height of the robot. Routes are calculated from one point to the next using the geometry of the robot while maintaining a safe distance from objects. In the case of multiple robots, combined routes are calculated for the controller and slave such that they remain in sight of each other as much as possible to enable tracking which results in better localization.
Peope are extremely good at encountering a new environment, quickly evaluating the conditions, making a plan and efficiently performing a task. Vision Robotics has several key innovations that enable its robots to perform at significantly greater speeds than other similar systems. In many applications, our systems operate in a two-step process. In the first step, the system models its environment and pre-plans the motions. This technique often decreases system costs while increasing operational speed and quality.
The basic approach is that the robot system explores, or maps, its environment before it begins its task. This time consuming mapping only needs to be performed one time. Once the map is complete, it is straightforward for the robot to execute its task. For example, a cleaning robot will map the house the first time it cleans. Each subsequent cleaning may be completed quickly because the basic map already exists. Pre-planning also enables efficient operation of complex tasks because the system has analyzed its moves prior to beginning so it is able to implement a complex plan.
This is the principle operation phase of the robot, during which the robot performs its designated task whether it is vacuuming the floor, patrolling an area or delivering parts in a factory. The algorithm begins with the valid perimeter map for the area, and may include additional exploration to re-map areas in which significant changes are detected. The actual algorithms are highly task and hardware specific and are developed in close collaboration with VRC's strategic partners.
When first started, the robot determines whether or not it has previously mapped the environment by searching its stored maps for features such as corners, doors, trees, rocks etc. The robot will begin its functional task if it recognizes the environment. Alternatively, it will explore and map the environment.