UAV Package Delivery Obstacle Avoidance issue

Hi, I am encountering an issue when trying to test the obstacle avoidance algorithm on a different scenario. I am using QGC connection for the trajectory and the quadcopter simply does not avoid all the obstacles, but only some of them. I've changed the safety distance, but without success. On QGroundControl, the quadcopter appears to "want" to avoid the obstacle, but it still follows the trajectory towards the landing point (see image)

I've tried to use a different obstacle but same thing happens - it avoided the red building, but went through the circular blue obstacle:

The lidar seems to be working fine. I would appreciate some help. Also, if I understood correctly from other questions on this forum, the project is designed for only one waypoint (i.e. take-off to landing)?

Many thanks in advance.

Tudor

Answer 1

[Jianxin Sun]

Hi Tudor,

Here are a couple examples that might help you:

https://www.mathworks.com/help/uav/ug/uav-obstacle-avoidance-in-simulink.html shows how to setup a Simulink model that navigates through multiple waypoints while avoiding obstacles

https://www.mathworks.com/help/uav/ug/tune-3d-vector-field-histogram-controller-for-obstacle-avoidance-in-3d-scene.html shows how to tune the obstacle avoidance algorithm with the help from MATLAB visualization

Thanks,

Jianxin

[Tudor-Stefan Tonca]

Hi @Jianxin Sun, many thanks for your reply.

I have checked the links, very useful towards obstacle avoidance validation.

However, is there a way through which more waypoints could be followed by the quadcopter within the UAV package delivery project? I'm looking into also adding high fidelity and recreating photorealistic environment, while maintaining the connection to QGroundControl. From QGC, I would like to add more waypoints than just the landing one. Currently, the quadcopter lands at the first waypoint.

Thanks

Tudor

[Jianxin Sun]

In order to navigate through multiple waypoints while performing obstacle avoidance, you probably need to combine the guidance logic (uavPackageDelivery/Multirotor (MultirotorModel)/Guidance Logic/Full Guidance Logic) with the obstacle avoidance controller (uavPackageDelivery/On Board Computer/DataProcessing/Process3DSensorData) as well as waypoint following controller (uavPackageDelivery/Multirotor (MultirotorModel)/Guidance Logic/Obstacle Avoidance3D).

It would look like:

QGC -> (Mission Points) -> Guidance Logic -> (From and To points) -> Obstacle Avoidance Controller -> (Target Point) -> Waypoint Controller

This will require some extensive change to the model though.

[Tudor-Stefan Tonca]

@Jianxin Sun, thank you very much for your answer.

[Tudor-Stefan Tonca]

@Jianxin Sun, I'm currently working on changing the model such as it allows multiple waypoints within the obstacle avoidance guidance logic. I'm thinking that I need to somehow combine the guidance logic from Full Guidance Logic with the one from Obstacle Avoidance (compute InnerLoop Cmds), as per your answer. However, the MATLAB function contains this code:

function y = computeLPWaypoints(mission, Steer,UAVState)
%#codegen

Wp1 = [mission(1).position' mission(1).params(4)];
Wp2x = mission(2).position(1);
Wp2y = mission(2).position(2);

if(~isnan(Steer))
    
    LAP_local = 6*[cos(Steer), sin(Steer)];
    LAP = [ LAP_local(1)+ UAVState.x , LAP_local(2) + UAVState.y , -11 , Steer];   
    if(isnan(LAP(4)))
        LAP(4) = 0;
    end
   
    y = [ Wp1(1,:) ; LAP ; Wp2x , Wp2y , 0 , Steer ];    
    DTD = sqrt( (LAP(1) - Wp2x)^2 + (LAP(2)-Wp2y)^2);  
    
else
    y = [ Wp1(1,:) ; UAVState.x , UAVState.y , UAVState.z-2 , 0 ; Wp2x , Wp2y , 0 , 0 ];
    DTD = sqrt( (UAVState.x - Wp2x)^2 + (UAVState.y-Wp2y)^2);
end

if(DTD<5)
    y = [ Wp1(1,:) ; Wp2x , Wp2y, -11 , 0 ; Wp2x , Wp2y , 0 , 0 ];
end


end

Are Wp1 and Wp2 the waypoints, therefore limiting the number of waypoints to 2? Would this code need to be changed to allow multiple waypoints?

I also noted that the GCSCmds signal which is sent to the waypoint follower has to be as per:

The Waypoints input accepts a set of waypoints as either an Nx3, Nx4, or Nx5 matrix, where N is the number of waypoints. Each row is an [x y z] position vector with optional yaw angle and transition radius for the UAV.

Many thanks.

Tudor