Octinion and Mechatronics 4.0 team navigate outdoor AGV using ultra wide band technology


Due to recent technological developments, robots and Autonomous Guided Vehicles (AGV) are  becoming more common, even outside industrial environments. An inspiring example is Octinion's prototype strawberry picking robot.

Octinion developed a prototype strawberry picking robot (http://octinion.com). This robot is able to navigate autonomously in a greenhouse, visually locate strawberries, pick and sort ripe strawberries without damaging them. Picking quality and speed are comparable to the ideal human picker's, but the robot exhibits advanced quality checking, which results in improved sorting.

The prototype strawberry picking robot by Octinion

In order to be effective, AGVs need a localisation system to retrieve information about their exact location as an input for their navigation algorithm. For the strawberry picking application, a cm-accuracy at an update rate of 10 Hz is required. Different technological solutions exist to achieve this, each with its own pros and cons.

  • The first set of solutions is based on GPS. Low-cost GPS modules are on the market. However, their accuracy is too limited for the intended AGV application. Differential GPS systems exhibit an improved accuracy, but their cost is prohibitive. Also, GPS signals tend to be blocked by the surrounding buildings, trees, etc.
  • Visual beacons form an alternative solution, which relies on optical sensors to localise the object with respect to calibrated visual features (beacons) in the surrounding environment. Accurate solutions can be achieved, but their performance can be strongly influenced by the environmental conditions, e.g. illumination and weather conditions (e.g. fog, snow).
  • Ultra Wide Band (UWB) communication technology offers a third solution. UWB is a technology for transmitting information spread over a large bandwidth. A two-way time-of-flight can be measured from communication exchanges between anchor and tag, which then can be used to calculate a range (see figure below). By triangulation, 3D coordinates of an object can be calculated. UWB looks like a promising alternative with the potential to achieve a 10 cm accuracy and hence to circumvent the disadvantages of both GPS and visual beacons. However, technical feasibility of Decawave UWB for this kind of applications is still unclear.

UWB communication to retrieve a ranging measurement

In this project Octinion and the Mechatronics 4.0 project team demonstrated the feasibility of Decawave Ultra Wide Band technology to achieve accurate localisation for navigation of an AGV. The architecture of the system is illustrated in the next figure. The AGV is equipped with a UWB tag and four anchors which are mounted on the edges of the operating range.

As the UWB system is only able to achieve an accuracy of 10 to 20 cm, its global position measurement is fused, using a Kalman Filter with signals originating from a low-cost Inertial Measurement Unit (IMU) and wheel encoders. This allows not only to achieve the required accuracy for the precise position control, but also to retrieve information regarding orientation and velocity of the AGV, which is required for its navigation. The IMU measures accelerations and rotational speeds in a local coordinate system. The scheme for this sensor fusion is illustrated in the following figure.

 Schematic representation of the UWB positioning system (left) and sensor fusion approach to achieve an accurate position and orientation estimate

The figure below shows the resulting output of the estimator (red) and compares it with the rough UWB data (purple). The noise level on the localisation estimate is significantly reduced. Stochastic analysis of the error shows that the standard deviation on the position measurement is around 50 mm, while the standard deviation of the orientation is about 1°.

Realised position estimate using sensor fusion of UWB localisation, IMU and encoder data 

Learn more on how the Mechatronics 4.0 team and Octinion achieved these results and witness an AGV navigating, based on UWB data during the Mechatronics 4.0 Masterclass on localisation on 28 March.

Contact persons Flanders Make: RisangGatot.Yudanto@flandersmake.be and Maarten.Witters@flandersmake.be