Jaco 2

This was the robot I spent the least hands-on time with (~6 months), but it was part of an interesting project: GANResilRob. My role was to design the integration layer connecting the perception and control modules being developed by other partners — basically the glue that lets everything talk to each other.

Getting hands-on with the real arm

Early on, I had the chance to set up the physical robot for tests and get familiar with the stack, mostly built around MoveIt. It was a nice opportunity to see how motion planning, collision checking, and trajectory execution come together in practice for a 7-DoF manipulator.

Moving to simulation

For most of the project, I worked with a Franka Emika Panda arm in Isaac Sim. The photorealistic rendering was a big draw, especially for any perception components that needed to be trained or tested on visually realistic data.

Working with the Panda was also a good opportunity to dig into some of the standard motion planners (e.g. OMPL-based planners available through MoveIt), compare Cartesian-space versus joint-space control, and get a feel for the challenges of planning in a 7-DoF arm. The extra degree of freedom compared to a more common 6-DoF arm means there’s a whole null-space of joint configurations that all reach the same end-effector pose, great for avoiding obstacles and joint limits, but it adds real complexity to the search space.

Some fun

As a bonus, I got the arm to serve me water. With a healthy dose of supervision and zero autonomy involved, but still a fun way to wrap up the project on a lighter note.