vamp Directory Reference

Directory dependency graph for vamp:

Files
	motion_benchmaker_demo.py
	MotionBenchmakerDemo.cpp
	MotionBenchmakerDemo.h
	MotionBenchmakerDemoMain.cpp
	Example program demonstrating how to use the Motion Benchmaker with OMPL and VAMP for evaluating motion planning on MBM problems.
	vamp_planning.py
	vamp_state_space.py
	VAMPPlanning.cpp

Detailed Description

VAMP-OMPL Demos

This directory includes demos using VAMP SIMD-accelerated state validity checks. VAMP uses robot-specific compiled forward kinematics operations and vectorized collision checking for state and motion validation.

C++

Geometric Planning

Run the simple geometric planning example from VAMPPlanning.cpp:

cd build/Release
cmake ../..
make -j <cores>
./demos/demo_VAMPPlanning

This will run a single plan instance on a spherized collision environment. To run OMPL's benchmarking on this, run:

./demos/demo_VAMPBenchmark --benchmark <number_of_trials>

This will run the benchmark for the specified number of trials with planners specified in VAMPPlanning.cpp and output the results in a .log file.

MotionBenchmaker

First, extract the json motionbenchmaker from the included VAMP submodule, from your ompl directory:

cd external/vamp/resources
python problem_tar_to_pkl_json.py

This will generate a file in <ompl_directory>/external/vamp/resources/panda/problems.json. Check the problem_tar_to_pkl_json.py script for more details and how to use with other robots.

Go back to your OMPL build directory and run the benchmark:

./demos/demo_VAMPMotionBenchmaker --problem-file <path_to_generated_JSON_file> --robot panda --planner RRTConnect --timeout 1.0

This will run the specified planner on all 100 problems for each 7 problem types from the MotionBenchmaker dataset. To run OMPL's benchmark:

./demos/demo_VAMPMotionBenchmaker --problem-file <path_to_generated_JSON_file> --robot panda --planner RRTConnect --timeout 1.0 --benchmark <number_of_trials>

This will run a benchmark for <number_of_trials> trials on one problem for each of the 7 MotionBenchmaker problem types, and output the results as a .log file.

Using log files from Benchmark

Use the provided Python script to analyze the generated .log files from the benchmark runs:

python <ompl_directory>/scripts/ompl_benchmark_statistics.py -d <generated_log_filename>.db <generated_log_filename>.log [<generated_log_filename2>.log ...]

These can be uploaded to PlannerArena for visualization

Python

Requirements

Install the required dependencies:

pip install requirements.txt

Geometric Planning

A simple script is provided to demonstrate the usage of VAMP with OMPL using Python bindings. Run the script from the demos subdirectory:

python3 vamp/VampPlanning.py

Run a benchmark with multiple planners and planning iterations:

python3 vamp/VampPlanning.py --benchmark --n_trials 100

If PlannerArena is installed, you can visualize at localhost:8888 or the port printed in the terminal output.

Visualize the robot, environment and planned trajectory at localhost:8080:

python3 vamp/VampPlanning.py --visualize

Motion Benchmaker Dataset

A script solving the 700 problems from 7 different problem sets on the MotionBenchmaker dataset. It supports Panda, UR5 and Fetch robots, changing the robot argument to panda, ur5 or fetch. The VAMP submodule at external/vamp is required. To run:

python motion_benchmaker_demo.py --robot panda --visualize

To run a benchmark with multiple planners, on one of the problems for each set:

python motion_benchmaker_demo.py --robot panda --benchmark --n-trials <n_trials>