Reliable greenhouse movement was tested on path changes, stopping and speed limits
Anyone assessing autonomous logistics in a greenhouse looks beyond whether a platform moves. A grower or technical decision maker wants to know what happens on concrete paths and pipe rails, during a path change and during an emergency stop. VDL ETG Projects therefore validated line vision software on an autonomous driving platform at Tomatoworld. Within the Handsfree Agrifood ecosystem, this test shows how measuring and adjusting help improve autonomous greenhouse logistics step by step.
Innovation package, use case and type test
Greenhouse Horticulture
After Harvest
Status: completed
Technical functionality
Broad knowledge question
What an autonomous platform must handle on concrete paths and pipe rails
Greenhouse horticulture companies want to know how an autonomous driving vehicle with camera sensors and an AI model responds to differences in concrete smoothness and pipe rail dimensions. The test therefore looked at driving on concrete, responding to pipe rails and braking at different speeds. This shows growers which movements have already been tested and which points matter when they discuss autonomous driving in their own greenhouse.
Approach
Path changes, emergency stop and speed tested separately
VDL ETG Projects tested an autonomous driving platform for systems that can move, sort, package and store harvested products. Testers measured the speed and duration of path changes, braking distance during an emergency stop and the safe speed on the concrete path. Using vision technology for rail detection, the robot drove autonomously from the first to the last rail. This showed where the platform kept its route, stopped and stayed within the agreed speed.
Objective
From a robotics promise to measurable driving behaviour
Technical decision makers can only use autonomous logistics when they can see how the platform drives, changes path, stops and corrects its position. The method set out 50 planned path changes, divided between left and right, with a maximum of two realignments per path change. During a realignment, the platform corrects its position to get back onto the route. Testers also measured the emergency stop and driving speed on the concrete path, so the results are linked to greenhouse movements that growers recognise.
Result and reflection
Growers can see how the platform changed path, stopped and limited speed within this validatio
Growers and technical decision makers can place the results for four movements next to their own greenhouse practice. They can see how the platform changed path, when it corrected its position, how short its stopping distance was and whether its speed on the concrete path stayed within the agreed limit. At the same time, the report states that post-harvest functionality, the data platform and integration in a greenhouse environment will be tested later. This allows readers to use this step when assessing autonomous greenhouse logistics, without reading it as a broad application yet.
Succeeded
Growers immediately see the limit of the result. Of the 50 planned path changes, 40 were carried out and all 40 fell within the predefined criteria. They can therefore assess the completed part as successful path changes within this test, not as proof for all planned path changes.
Technical decision makers can use the realignment figure to see how often correction was needed. Of the path changes carried out, 58 percent succeeded without realignment and 42 percent with one realignment. This means that in 42 percent of the completed path changes, the platform corrected its position once to complete the path change.
Growers can link stopping behaviour to a concrete distance. Five emergency stops each remained within half a robot length at 1 m/s. Within this test, that shows them how short the stopping distance was at that starting speed.
Technical decision makers can compare the concrete path speed with the agreed limit. The platform drove at a maximum of 0.3 m/s and did not exceed the speed limit. This shows them that speed remained limited within this part of the validation.
Learning points
Readers also see which part was adjusted after the test. The engineering team saw that the alternative IMU was more accurate than the old IMU and therefore replaced it.
Because 40 of the 50 planned path changes were carried out, readers should read the positive result as the outcome of the completed part of the validation.