Issues with jacobi-motion-0.0.33

Unfortunately, I don't have much info. I can try to get back to you later (but I'll be on vacation very soon.)
From what I remember, I'm 95% sure it was a linear motion though, and that the amplitude of the motion itself was quite small (a few centimeters).
Hopefully, that helps a bit.

Some more details after I was able to reproduce the problem in a minimal script (see below):

• It happens only with LinearMotion
• Surprisingly, it seems to happen only for very specific timestep sizes:
  

  timestep: 0.005 - duplicate delta time: 0.555
  timestep: 0.01 - duplicate delta time: 0.56
  timestep: 0.02 - duplicate delta time: 0.56
  timestep: 0.023 - duplicate delta time: 0.552
  timestep: 0.031 - duplicate delta time: 0.558
  timestep: 0.04 - duplicate delta time: 0.56
  timestep: 0.041 - duplicate delta time: 0.5740000000000001
  timestep: 0.042 - duplicate delta time: 0.5880000000000001
  timestep: 0.046 - duplicate delta time: 0.552
  timestep: 0.047 - duplicate delta time: 0.5640000000000001
  timestep: 0.048 - duplicate delta time: 0.5760000000000001
  timestep: 0.049 - duplicate delta time: 0.5880000000000001
  timestep: 0.05 - duplicate delta time: 0.6000000000000001
  timestep: 0.062 - duplicate delta time: 0.558
  timestep: 0.079 - duplicate delta time: 0.553
  timestep: 0.08 - duplicate delta time: 0.56
  timestep: 0.081 - duplicate delta time: 0.5670000000000001
  timestep: 0.082 - duplicate delta time: 0.5740000000000001
  timestep: 0.083 - duplicate delta time: 0.5810000000000001
  timestep: 0.084 - duplicate delta time: 0.5880000000000001
  timestep: 0.085 - duplicate delta time: 0.5950000000000001
  timestep: 0.08600000000000001 - duplicate delta time: 0.6020000000000001
  timestep: 0.08700000000000001 - duplicate delta time: 0.6090000000000001
  timestep: 0.092 - duplicate delta time: 0.552
  timestep: 0.093 - duplicate delta time: 0.558
  timestep: 0.094 - duplicate delta time: 0.5640000000000001
  timestep: 0.095 - duplicate delta time: 0.5700000000000001
  timestep: 0.096 - duplicate delta time: 0.5760000000000001
  timestep: 0.097 - duplicate delta time: 0.5820000000000001
  timestep: 0.098 - duplicate delta time: 0.5880000000000001
  timestep: 0.099 - duplicate delta time: 0.5940000000000001

• I haven't seen more than 1 duplicate in the trajectory I chose
• This happens both with cloud and on-prem version of the lib (v0.0.32)

import jacobi

robot = jacobi.robots.UniversalUR5e()

joint_names = [
    "shoulder_pan_joint",
    "shoulder_lift_joint",
    "elbow_joint",
    "wrist_1_joint",
    "wrist_2_joint",
    "wrist_3_joint",
]

robot.set_base(jacobi.Frame.from_quaternion(x=-0.1, y=0.0, z=0.778, qx=0.0, qy=0.0, qz=0.0, qw=1.0))
robot.flange_to_tcp = jacobi.Frame.from_quaternion(
    x=-4.0155031343329083e-13,
    y=-0.08200070709142922,
    z=0.20374969879588292,
    qx=-4.496334640197861e-18,
    qy=-1.836602551794397e-06,
    qz=0.9999999999983135,
    qw=2.448180546764611e-12,
)

robot.max_velocity = [0.1 * v for v in robot.max_velocity]
robot.max_acceleration = [0.1 * v for v in robot.max_acceleration]

start = jacobi.Waypoint(
    position=[
        -0.2274138063906141,
        -1.224123650017857,
        1.9087685108689072,
        -2.254713396909295,
        -1.5737379891516146,
        -4.094896483092814,
    ]
)
goal = jacobi.CartesianWaypoint(
    jacobi.Frame.from_quaternion(
        x=0.4017176276920228,
        y=-0.048692228485077385,
        z=0.8390208291756214,
        qx=0.9131367110202927,
        qy=-0.40764452677770674,
        qz=8.542422724806657e-05,
        qw=0.0026980507637731866,
    )
)

scene = jacobi.Environment(robot=robot, safety_margin=0.0)

jacobi_planner = jacobi.Planner(scene, delta_time=0.02)

desired_motion = jacobi.LinearMotion("my_motion", robot, start=start, goal=goal)

trajectory = jacobi_planner.plan(desired_motion)

for t, p in zip(trajectory.times, trajectory.positions):
    print(t, p)

for dt1, dt2 in zip(trajectory.times[:-1], trajectory.times[1:]):
    if dt1 == dt2:
        print("duplicate delta time:", dt1)

@Alban Laflaquiere:

Fantastic! Thanks for all this information!

1. Looks like we failed to update the documentation with regards to the changes on that function. Sorry about that!

   set_base

   has been changed to

   base

   as a property, and would need to be changed as such in the code.

As for 2.

Looks like that indeed is a bug! Seems that we have tuned one of our internal parameters a bit too much which is causing some joint positions to duplicate due to generated waypoints being too close together. As I understand this shouldn't effect normal operation, and we will have a fix for it in next patch!

@Alban Laflaquiere: with regards to the

set_base

base