30 October 2023 Research Development Meeting

[Woolfrey]

Chair:

• Jon @Woolfrey

Attendees:

• Andrea @andrearosasco
• Stefano @steb6
• Vignesh @vigisushrutha23
• Simone @SimoneMic

Absent:

• Nicola @xEnVrE

---

Agenda:

Note: Last week's items should become this week's agenda.

Item	Tag	Responsible	Comment
Finished behaviour tree for manipulation from navigation	N/A	@vigisushrutha23	In progress; need to test in simulation.
Build docker for navigation & manipulation	N/A	@vigisushrutha23	Almost complete; tested and working in Gazebo.
Implement mobile robot trajectory in C++	N/A	@Woolfrey	Still in progress; hopefully this week it's works
Test vision and BT in Gazebo with newest manipulation code and push to demo branch	#133	@andrearosasco @steb6 @Woolfrey	We could possible test on ergoCub 1.1 before it leaves, but is currently under repair. Vignesh working on docker.
Develop pre-demo checklist	#132	@Woolfrey @vigisushrutha23 @steb6 @andrearosasco	Need to organise a meeting to finalize it.

Additional Discussion Points:

Anything not captured in the agenda.

Discussion Point	Raised By	Tag	Outcome
Obtained results for trajectory optimisation of mobile robots for use with MPC code	@Woolfrey	N/A	See discussion below.
We ran a brief "root-cause analysis" workshop to make sure future demos work	@Woolfrey	#131	Results posted in minutes of last week's meeting. New issues created to be worked on.
Working on MPC paper (both article and code)	@Woolfrey	N/A	See discussion below. Hoping to have numerical results as "proof-of-concept" by the end of November.
Workshop on Wednesday for planning end-of-year demo.	@Woolfrey	N/A	New project has been created. Meeting room booked for Wednesday November 2nd.
May have found a way to embed closed-loop dynamics in to a position controller with Verlet integration	@Woolfrey	N/A	See discussion below.

---

Next Meeting:

Note: These will become the agenda for the next meeting.

Item	Tag	Responsible
Organise meeting to develop pre-demo checklist.	#132	@Woolfrey
Add release tags to working code for demos	N/A	@Woolfrey @SimoneMic @vigisushrutha23
Upload experiment videos to HSP sharepoint	N/A	@vigisushrutha23

[Woolfrey]

Trajectory Optimisation for Wheeled Mobile Robots

Full details in article : Trajectory_Optimisation_for_Mobile_Robots-1.pdf

Some brief results for a trajectory where a robot must turn around and drive to the point (-1,1) in Cartesian coordinates:

[Woolfrey]

Model Predictive Control paper

Working on draft (need to share Overleaf project with Lorenzo Natale):

Simultaneously working on code. Need to implement:

1. Cubic polynomial class
2. Use polynomial to generate trajectories for wheeled mobile robot
3. Sample the trajectory for use in the MPC class.

[Woolfrey]

Dynamics in Position Control with Verlet Integration?

At the moment, I am using a velocity controller with discrete time integration to send position references to the robot:

Using the real joint values in the feedback makes the control chatter / unstable.

Standard Euler integration updates position, then velocity using discrete time steps:

BUT, it is known to be unstable for conservative systems.

Verlet integration uses position only, and is known to be stable:

It is theoretically possible to do dynamic control with Euler integration. The system should be controllable:

But using a Verlet integrator is also controllable:

It should be possible to emulate torque control in a position-controlled robot.