Inverse dynamics block to forward dynamics block

Hello, I am currently facing an issue related to the joint positions of a 6-axis robotic arm. My objective is to visualize these joint accelerations without using a controller(That will be done later). Here's my approach: I generate a trajectory using a minimum jerk polynomial trajectory, obtaining joint positions (q), velocities (qd), and accelerations (qdd). I then use these values along with an external force (fexterior) as inputs for the Inverse Dynamics block, resulting in joint torques.

Now, with these obtained joint torques, along with the initial joint positions (q), velocities (qd), and the external force (fext) used earlier, I attempt to regenerate the joint positions using the Forward Dynamics block. However, I observe that the joint accelerations remain constant even when I vary the external force (Fext). This is perplexing, especially considering that when I use the Inverse Dynamics block with different Fext values, I obtain distinct joint torques for each payload. Yet, the results from the Forward Dynamics block remain the same.

Could you provide insights into why the joint accelerations from the Forward Dynamics block are not varying with changes in the external force, despite the distinct joint torques obtained from the Inverse Dynamics block for different payloads? Your assistance is highly appreciated. Thank you!

Answer 1

[Karsh Tharyani]

Hi Denizhan,

Let's use the following as a starter code

% A 6 axis robot
rbt=loadrobot("abbirb120",DataFormat="column");

% Assume a gravitational acceleration in the Z direction of the Base frame.
rbt.Gravity=[0,0,-9.81];

% Generate a minimum jerk profile between some random waypoints and time
% points.
s=rng(100);
nwpts=10;
wpts=rand(nwpts,6);
tpts=linspace(0,1,nwpts);
[q,qd,qdd]=minjerkpolytraj(wpts',tpts,100);


% Inverse dynamics on the first set of joint states. Additionally, assume 
% a non-zero external force acting on the end-effector body "tool0".
FIRST_JOINT_STATE_IDX=1;
Q=q(:,FIRST_JOINT_STATE_IDX);
QD=qd(:,FIRST_JOINT_STATE_IDX);
QDD=qdd(:,FIRST_JOINT_STATE_IDX);
eename='tool0';
fext=externalForce(rbt,eename,rand(1,6),Q);
tau=inverseDynamics(rbt,Q,QD,QDD,fext);
disp(tau)

    1.0241
  -17.3311
   -3.7767
   -0.8281
    0.3237
   -0.3762

% Feed this system state to the forward dynamics function
QDD_calc=forwardDynamics(rbt,Q,QD,tau,fext);
disp(QDD_calc)

     0
     0
     0
     0
     0
     0

Do you see why QDD_calc is zero? It is zero because the joint torque tau along with the joint state Q, QD and the external force fext balances the system.

Let me know if you have any questions or if I didn't get your question correctly.

Best,

Karsh

[Denizhan AKINCI]

I'm sorry for the delay in responding. Your help was greatly appreciated, thank you.

Regarding the acquisition of torques, is it reasonable to attempt regenerating accelerations and velocities without a controller or its corresponding code? When evaluating the performance of a robotic arm prior to building the actual controller, is there a way to assess its dynamic capabilities without considering the controller component?

My supervisor was the one that told me to use the method I mentioned above (converting joint positions, velocities, and accelerations to torques, then using those torques along with joint positions and velocities to derive accelerations, and finally integrating accelerations to get velocities) to evaluate the robot's capability to complete tasks. Also see which joints exceeds their limits etc. By the way my supervisor is not a robotics expert at all, we were just trying mehods.