Motion Planning
2021.8.17
Paper – Motion Planning with Sequential Convex Optimization and Convex Collision Checking
- sequential convex optimization
- convex collision checking
- CHOMP (Covariant Hamiltonian Optimization for Motion Planning)
- At the core of our approach are (a) a sequential convex optimization procedure, which penalizes collisions with a hinge loss and increases the penalty coefficients in an outer loop as necessary, and (b) an efficient formulation of the no-collisions constraint that directly considers continuous-time safety
- statically stable walking motion
- We also apply TrajOpt to plan curvature-constrained steerable needle trajectories in the SE( 3) configuration space and multiple non-intersecting curved channels within 3D-printed implants for intracavitary brachytherapy.
- revived
- non-holonomic
- swept-out volume
- polyhedral approximation
- quantitative
- simultaneously
- torso
- optimize clinically relevant metrics for flexible, bevel-tip medical needles
- multiple bevel-tip flexible needles inserted through the perineum to reach targets deep within the prostate following high-quality constant curvature trajectories, and (f) optimized layout for bounded curvature channels within 3D-printed vaginal implants for delivering radiation to OB/GYN tumors.
- (1) we use a different approach for collision detection, and (2) we use a different numerical optimization scheme.
- Two key ingredients of a sequential convex optimization algorithm are: (a) a method for constraining the step to be small, so the solution vector remains within the region where the approximations are valid; (b) a strategy for turning the infeasible constraints into penalties, which eventually drives all of the constraint violations to zero.
- Our implementation uses a variant of the classic 1 penalty method (Nocedal and Wright, 1999), described in Algorithm 1
2021.8.20
Motion Group
2021.8.18
- Moper: Motion planner package includes collision check and trajopt — TOPPRA
- xyz-urdf: provide DH parameters and mesh
- xyz-kdl: FK/IK/MoveL/MoveJ — TOPPRA
- xyz-motion-base: includes xyz-robot and xyz-kdl… Python interface ---- TOPPRA
- piece picking has no collision check !!!
- movement flow is in RAFCON, which includes waypoints to ensure trajopt.