2020 AGV Industry Trend Forecast: Vision Navigation Facing Upcoming Exposure?

2020 AGV Industry Trend Forecast: Vision Navigation Facing Upcoming Exposure?

Author: Huang Manting from xzlrobot.com

In terms of technological development and end user needs, vision navigation, as an emerging navigation method in AGV industry, is widely considered to be AGV industry mainstream in the future, and is expected to be a breaking dark horse in 2020. Taking the example of VisionNav Robotics, a representative enterprise in vision navigation AGV. It takes only 3 years to achieve turnover of more than 50 million Yuan in the sales of unmanned industrial vehicle, which attracts a lot of attention.

Table 1 Advantage and Disadvantage of Different Navigation Types

Navigation Type	Advantage	Disadvantage
Magnetic Navigation	Accurate positioning, relatively easy to place, change or extend routes, comparing to electromagnetic navigation, low cost	Vehicle can only run along with magnetic tape, cannot change tasks in real-time, magnetic tapes are easy to worn and need regular maintenance.
QR-code Navigation	Accurate positioning, small size with flexibility, relatively easy to place and change routes, communication easy to control, no interference to sound or light.	Routes need regular maintenance, and if it’s in complex sites, need to replace QR code label constantly.
Laser Navigation	Mature technology, enable to plan route flexibly, high positioning accuracy, flexible driving route, relatively easy for installation.	Manufacturing cost and price are relatively high, detective range is limited, mainly applied in indoor environment; missing semantic information in the map limits the ability to expand application in complex environment.
Vision Navigation	No limits for sensor detection distance, flexible route planning, high positioning accuracy, easy for installation, able to work both indoor and outdoor environment, low cost, able to extract semantic information in the map, high reusability in complex environment.	High dependency on the light, auxiliary sensors are needed in dark places or texture-free areas

Vision Navigation Basic Categories and Characteristics

Take sensor as the example, there are mainly 3 ways to realize the routes in Visual SLAM solutions: RGB-D depth-cameras such as Kinect/Kinect V2、Xtio; monocular (fisheye) cameras; binocular (or multiocular) cameras. In terms of realization difficulties, the general order is monocular vision > binocular vision > RGB-D.

Monocular SLAM is also called MonoSLAM，who can complete SLAM with only one camera. The benefits are that sensors are simple with low cost, therefore MonoSLAM has attracted much attention from researchers. Compared to other visual sensors, there is no need for monocular cameras to obtain environment absolute depth, so MonoSLAM is not affected by the size of the environment, and can be used in both indoor and outdoor.

In contrast to monocular, binocular (multiocular) cameras estimate spatial points locations through the baseline among multiple cameras, it can also estimate the depth of the environment  both in motion and motionless. Different from monocular, the configuration and calibration of binocular (multiocular) camera is relatively complicated, and its depth measurement range is limited by the baseline and resolution of the binocular, so FPGA is required to complete distance measurement.

RGB-D camera started to rise around 2010. Its biggest characteristics is that it can directly measure the distance between each pixel in the image and the camera through infrared structured light or Time-of-Flight principle. Therefore, it can provide more information than traditional cameras, and it can spend less time and efforts than monocular or binocular to calculate the depth. However, most RGB-D cameras still have problems such as narrow measurement range, large noise, and narrow views. Due to limit measurement range, it is mainly used for indoor SLAM.

Development Process of AGV Companies Applying Visual Navigation Technology

With continuous development of Industry 4.0, industrial automation and intelligent upgrades have more requirements to AGV for higher accuracy. At present, positioning accuracy of vision system used in automated production lines can reach 0.1mm. But for pure visual navigation AGV without fixed reference, its positioning accuracy is about 10mm, yet to be improved. In order to obtain higher accuracy, in addition to visual technology, AGV vision navigation systems also need to be combined with the assistance of other sensors.

How to improve positioning accuracy is always a problem that all AGV companies have been exploring. At 2019 CeMAT Asia, VisionNav Robotics unveiled their vision navigation control module stunningly. This universal module, which took 2 years to build, enables unmanned industrial vehicles to achieve positioning and operation accuracy of ± 3mm. and has been given its world’s first show in a precise multi-layer stacking solution based on vision technology. "In our industry, all valuable technologies and products need to be precipitated and excavated the requirement and application scenarios from a large number of customers, and in turn to force developer, and verify development results in customer’s actual requirement and application scenarios, and then forced developer again, to realize continuous iteration. If you only make a laboratory prototype or have just deployed 1-2 customer sites, you still have a long way to go for mass production, because you may not even understand customer’s actual requirements " said Dr. Luyang Li, CEO of VisionNav Robotics.

The responsibility of leading enterprises in the industry is to promote the appearance of industry explosion breaking point. Whether there is explosion breaking point for vision navigation in the fields of unmanned industrial vehicle, Dr. Li Luyang believes, the decisive factor is not market demands, but technology and product. “It is fair enough to say that there are no any unmanned industrial vehicles of any companies worldwide can fully meet the customer's requirement on efficiency, function, cost, and vehicle reformation. Therefore we will not feel any comfort for the current situation, instead we should increase the investment on R&D, strive forward with our whole heart and mind and never stop working deeply on vision navigation and products until our technology and project can fully meet customer’s requirement.“

Core Advantage of Vision Navigation at Current Stage

After several iterations of vision navigating AGV, there is a great improvement for its stability and reliability. Mature vision navigation has the following three core advantages:

1）Strong Usability: Compared to other navigation types, vision navigation can extract semantic information from the environment, therefore it can adapt to complex scenarios in the demand of customer’s regular use. For example, deformed racks or pallets after long-time use, large proportion of oversize goods, unpredictable transportation needs with periodical surge, vehicle dispatching and scheduling in narrow space, etc. Take VisionNav Robotics, who have several pioneering solutions for complex scenarios in the industry, as an example. Vision navigation not only enables AGV to firstly realize unmanned high level shelf operation at 9.4m, but also creatively solves unmanned transportation problem for multiple irregular parts combining industry-leading self-adaptive picking and placing goods technology and real-time positioning error correction technology,

2）Wide applicability: Adding modularized vision navigation technology to industrial vehicles can easily achieve unmanned transformation and upgrade. At present, vision navigation module from vision navigation AGV companies, such as VisionNav Robotics, can adapt to over 180 industrial vehicles models from 6 major brands.

3）Cost Effective: Cost advantage of vision sensors has significantly reduced the price of the product. The price of industrial camera from world’s top-level supplier is around 3000 RMB, which is 10%-20% of the price of same level laser sensors. Therefore, in terms of ROI or payback period for end customer, there is an obvious advantage for vision navigation unmanned industrial vehicles.

Dr. Li Luyang, co-founder of VisionNav Robotic told reporter from xzlrobot.com in an interview earlier, that “VisionNav Robotics is still working on vision navigation technology at the current stage. Recently we have reached a 5G strategic cooperation fram agreement with ZTE, in order to promte the combination betwwen 5G technology and vision navigation industrial unmanned vehicles, and push forward the landing of innovation application scenarios and solution research. We have confidence to combine vision and the motion of industrial vehicle in a more perfect way in 2020, further improvet the product usability, applicability and cost performance, and promoting the appearance of industry break point.