PlanningService API

The robot uses a cd to solve for motion plans which satisfy constraints. This PlanningService consists of an app which: - runs a loop that has a comms layer listening for messages and publishing to the client which provides an abstraction to many different communication paradigms (lcm, ROS, Viam) - a planner that tranlates the PlanningProblemDefinition into a set of drake::Constraints and solves the planning problem - a data layer which stores pre-computed information to reduce planning times, and periodically makes insertions into the database to improve performance - a client interface which inherits from the same comms library used by the comms layer and allows other code to use the service.

It is also possible to directly instantiate a planner in c++ code and not make use of the services architecture.

The user transmits a robot_description (either URDF or SDF), a PlanningProblemDefinition which consists of a start, a goal, and a set of rules for creating a set of drake::Constraints via the communication_layer which then populates a c++ struct and calls the planner. The PlanningResult is serialized by the communication_layer and transmitted back to the caller.

  graph LR;
    A[client] -- Serialize --> B[comms];
    B -- c++ api --> C[planner];
    C --> B;
    B -- Serialize --> A;

Supported Functions

We support two different methods for finding a motion plan. Either arbitrary motion is allowed along as it satisfies constraints, or we require the specific motion of a frame and solve via IK.

// these are really the same
FindMotionPlan();
CreateActionPart();
ConstrainedMotionPlan();

FindLinearMotionPlan();
IkMotionPlan();
InverseKinematicsMotionPlan();
LinearMotionPlan();
FrameConstrainedToFunctionMotionPlan();

Developing

Prerequisites

docker and the docker compose plugin for docker

Overview

The planning service is built with bazel inside a docker image. There are two ways to run it 1. bazelw is an included wrapper that will run all of the bazel commands inside the image with appropriate mount points and configuration for a development setup. This is easiest to get setup. 2. dazel we are experimenting with dazel. To run with dazel clone https://github.com/nadirizr/dazel to your local system and install the dazel script somewhere on your local path, e.g. on linux:

cd $HOME
git clone git@github.com:nadirizr/dazel.git
cd dazel
pip install dazel
sudo ln -s $HOME/dazel/scripts/dazel /usr/local/bin/dazel

Dazel may be configured with the .dazelrc file.

Common Tasks

./run_unit_tests.sh to build and test your changes docker compose build to rebuild the image in case you have made Dockerfile changes

Documentation

./scripts/generate_docs.sh to build and view doxygen-style documentation.