GreenDigger Collaboration

Autonomous Mobile Platform

Mobilising a platform on a hardpan environment so rainwater can efficiently be caught and used by the soil.

Object Recognition

Using OpenCV functions the image recorded by a depth camera is analysed and used for object detection. This later informs the pathfinding algorithm.

OpenCV Stack

First the image received from the depthcamera is read. This image is in grayscale and makes it very capable for a Canny edge filter. Using the edited image, bounding boxes are drawn based on the now amplified edges.

Distance

After the bounding boxes are drawn a calculation can be done based on the view of the camera where the next obstacle is. The result of this calculation is a vector to the next object. Adding this to the position of the AMP at that moment, results in the real coordinates of the obstacle.

Pathfinding

To drive autonously a route should be determined. In the previous section possible hinderances could be determined. Now navigating around them is of the order.

Why and How?

GreenDigger supplied a map of where the planned excavation spots would be. Using this, the next goal could be determined and a simple pathfinding algorithm could be chosen where one start position and one goal was used. Since we know the position of the goal, the heuristicly informed A* was chosen. This is a fast algorithm which uses heuristic function to plan the best route.

Driving

Now that a path has been generated by the A* algorithm we are able to drive the path it created. This is done by first turning towards the next vertex and then driving there. This method is chosen because as little errors as possible can be made using this approach and the machine will not drive large distances.