Collision avoidance sensors for flying in dense forest

Hi,

I'm trying to design a collision avoidance system for flying a hexacopter with a 1.5m diameter through a dense plantation of areca nut trees. There is about a 5m space between the trees. The objective is spraying the areca nuts.

I'm trying to determine the suitable sensors that could be used in the collision avoidance system to prevent the drone from damaging the leaves and branches while flying between the trees under the forest canopy (and also from flying into the trees ofcourse).

I'm thinking 360 LiDAR and TOF cameras (flash LiDAR), but I don't know if these would be able to detect leaves and branches.

Could anyone advise me on the range sensors that could be used for collision avoidance when flying under a forest canopy?

You need to be a member of diydrones to add comments!

Join diydrones

Email me when people reply –

Replies

  • Thank you Laser Developer. It looks like I will be contacting the ArduPilot team and some lidar vendors soon.

  • @Khushal B - LiDAR systems can work at close and far range equally well. We test the SF40 down at 1m and it is still within spec. Below 1m the accuracy may deteriorate slightly but it still gives useable results.

    The issue of FOV is always a tricky one. We think that conical (circular or elliptical) scanning patterns have great potential for obstacle detection but there are no flight controllers currently supporting this mode of operation. The next version of the LW20 will support conical scanning.

    At this time there is limited support for collision avoidance on all autopilots. We are hoping to spend more time on this later in the year after we have completed some other projects but for your application you would be looking at developing special code. The ArduPilot team is very helpful in this regard and will do contract development for you if you don't want to do it yourself. We are one of their sponsoring companies and they are familiar with our products but even if you decide to use other technology I would recommend making use of their services.

  • @Laser Developer - Thank you for that insightful reply. I had a look at the range of products offered by LightWare, including the use of SF11 by Cala as a laser altimeter. I was planning to use an ultrasonic sensor for the altimeter, but I'm considering using this lidar now instead. I also came across this: http://ardupilot.org/rover/docs/common-lightware-sf40c-objectavoida.... ArduPilot supports the LightWare SF40c 360degree lidar, but will it be as effective for very short range under 1m as well? I'm looking for obstacle avoidance solutions when flying in tight spaces between trees and branches.

    Would a single lidar have enough fov to be used for obstacle avoidance on top of the hexacopter?

    Also, ArduPilot currently supports object avoidance only in the direction the vehicle is travelling, so I could end up scraping trees or branches on the sides. Any ideas on how I could get around this?

    Thanks, and sorry for the late reply.

    LightWare SF40/C 360 Lidar — Rover documentation
  • @Khushal B - In our experience, LiDAR gives good results when measuring to foliage, even against low density leaves on trees and bushes. Using a spinning LiDAR gives great data coverage in one plain but acting on the data for collision avoidance is not yet a standard capability for most flight controllers. TOF cameras tend to be limited in range, especially in bright sunlight conditions and I don't know anyone who has used this technology successfully in a drone application.

    As an alternative, the very small and low cost LW20 LiDAR includes a servo driver that makes it easy to run an automatic scanning pattern in the forwards looking direction or to manually direct the LiDAR to aim in a specific direction.

    There is a two axis version of the LW20 servo driver coming out next year that allows for 3D mapping over a limited field of view. The unit can be directed to look up or down, side to side in a sweep, make a 3D map, perform conical scanning (the most efficient way to find tree trunks) or move in random 3D patterns. L.D.

This reply was deleted.

Activity