New-Tech Europe Magazine | Q1 2020

SLAM SLAMisquicklybecomingan important advancement in embedded vision as it enables a device with the ability of location awareness. Using SLAM, a vehicle can not only track where it is heading or its direction (orientation), but also how it is moving within its surrounding environment (location, speed and altitude). Computations for SLAM were typically performed with a camera sensor as the only form of input. This was known as Visual SLAM (VSLAM). But in the past few years, with the suite of additional sensors becoming available, SLAM has evolved to fusing additional sensor inputs. A SLAM system works by tracking a set of points through successive camera frames and other sensor data to triangulate the camera’s 3D position, while simultaneously using this information to approximate camera (or another sensor) orientation. As long as there is a sufficient number of points being tracked through each frame, both the orientation of the sensor(s) and the structure of the surrounding physical environment can be rapidly understood. So, for example, in the case of automotive implementations, SLAM may use a combination of one or more forward-facing cameras, radar, lidar, and inertial measurement units (IMUs, which provide data from accelerometers and gyroscopes that help to estimate the sensor’s orientation) as inputs. SLAM is then used to determine how the vehicle is moving in the environment. When GPS data is available, it can be used to fortify the position estimate. Figure 2 shows an example in which a variety of sensors, such as camera, lidar, and radar, is mounted around the vehicle that can be used as input for SLAM. SLAM applications SLAM is a key ingredient in many

applications that are used for driver assistance and self-driving vehicles. A few of these applications include: Lane Keeping Assistance (and Lane Departure Warning): In addition to tracking lane markings on the road, SLAM is used to make sure the vehicle traveling safely within a lane and to engage in lane changes safely. Navigation: By understanding the surrounding environment combined with a planned route and GPS data, the vehicle can use SLAM to pilot itself to its destination. Forward Collision Warning (FCW): Combined with SLAM, the path or trajectory of the current vehicle can be used to for more robust collision warning. Market trends for SLAM As shown in Figure 3, the market size for SLAM-based applications is set to exceed $2 billion by 2024 [1]. Major drivers for this market growth are the advancements in SLAM algorithms and the growth of SLAM in various markets. The rising technological developments and growing awareness regarding the benefits offered by SLAM are primarily driving the market demand.

Figure 1: 6DoF

Growing interest in the technology, particularly from industries including autonomous vehicles and augmented virtual reality, has resulted in the adoption and expansion of SLAM across the globe. Moreover, SLAM used for navigation in both indoor and outdoor environment applications opens an opportunity for the larger adoption of the technology across various end- user industries. Over the past five years, leading technology companies have made significant investments in SLAM to integrate into various business expansion strategies such as new product developments and mergers and acquisitions.

Figure 2: AVs use many sensors and cameras to perceive their surroundings

New-Tech Magazine Europe l 21