Final work proposals

Standard proposal

The objective of the final work is to put into practice all the theoretical concepts on motion and task planning studied through the tutorials but using high-DOF robotic systems like:

  • Mobile manipulators (Tiago or the TX-90 with mobile base)

  • Dual-arm robots (Yumi demos or the Chess example with two UR3 arms).

  • Multi-robot systems composed of several 2-DOF robots.

  • Other robotic systems involving mechanical hands.

The use of challenging scenarios different from those available is encouraged.

The work has two parts:

  • Motion planning: the evaluation of different motion planners using the benchmarking utility in qualitatively different queries.

  • Task planning: the setting of a task and motion planning instance through the definition of the PDDL domain and problem files, and the task configuration file.

The deliverable should include a pdf file describing the work done and a video with the execution of the task and motion planning problem.

Proposal with a theoretical bias

Complementarily to the standard proposal, a proposal with a more theoretical bias can be chosen among the following two options. Both of them include a theoretical part plus a simple version of the standard proposal. The objective is on the one hand to deepen into more advanced motion planners, and on the other to put into practice the concepts learned in some simple scenarios like those used in the tutorials.

Proposal A

  1. A theoretical study of one of the following planners available in Kautham and its evaluation and benchmarking against the planners used in the tutorials.

  • The Vector Field RRT

  • The LazyRRT planner

  • The RRT and SST planners to plan with kinematic constraints

  • The RRT* variant for high dimensional configuration spaces

  1. A simple version of the standard proposal

Proposal B

  1. A theoretical study of one of the following four advanced planning approaches based on some recent published papers:

  • Constraint-based motion planning

  • Optimization-based motion planning approaches

  • The LTL-based planners

  • Physics-based motion planners

(No planners of these types are available in Kautham and therefore this is a pure theoretical study).

  1. A simple version of the standard proposal

Proposal with an implementation bias

Alternatively, a pure practical proposal is possible for those students aiming to contribute to The Kautham Project by e.g.:

  • Modifying the RRT planner to incorporate the dynamic domains.

  • Adding a new OMPL planner (RRT#, the Informed RRT*,…)

The evaluation of the planner implemented using benchmarking is desirable.