README.md

Planning Heros

Planning and Decision Making for Autonomous Robots Project, Fall 2021 @ETHZ

Installation

Required python version:

Item	Value
Required Python Version	3.9
Deadline	31.12.2021, 23:59:59

Running from the console, assuming you run this from the project root:

# install requirements
$ pip install -r requirements.txt
# export PYTHONPATH to run from console
$ export PYTHONPATH="$PWD/src"
# run app_main.py
$ python src/pdm4ar/app_main.py

Running from the provided Docker container (Requires a recent docker version to be installed).

$ make run-final21

Generated Solutions

Below you can find some visualizations of the scenarios we used to evaluate our approach.

Static Scenario:

EpisodeOutcome(
	goal_reached        =	    1,
	has_collided        =	    0,
	distance_travelled  =	133.38,
	episode_duration    =	10.68,
	actuation_effort    =	11.03,
	max_acc_lat         =	 9.63,
	max_acc_long        =	23.56,
	avg_computation_time=	 0.10,

)

Dynamic Scenario:

EpisodeOutcome(
	goal_reached        =	    1,
	has_collided        =	    0,
	distance_travelled  =	136.09,
	episode_duration    =	12.30,
	actuation_effort    =	10.77,
	max_acc_lat         =	 9.63,
	max_acc_long        =	20.21,
	avg_computation_time=	 0.36,

)

Custom Dynamic Scenario 1:

EpisodeOutcome(
	goal_reached        =	    1,
	has_collided        =	    0,
	distance_travelled  =	134.12,
	episode_duration    =	 9.57,
	actuation_effort    =	11.25,
	max_acc_lat         =	11.65,
	max_acc_long        =	22.88,
	avg_computation_time=	 0.55,

)

Custom Dynamic Scenario 2:

EpisodeOutcome(
	goal_reached        =	    1,
	has_collided        =	    0,
	distance_travelled  =	145.94,
	episode_duration    =	16.29,
	actuation_effort    =	 8.31,
	max_acc_lat         =	42.71,
	max_acc_long        =	42.71,
	avg_computation_time=	 1.18,

)

Solution approach

• RRT* - run for each timestep:
  • Create Motion Primitives
  • KD-tree to compute distances
  • Caution, use box-distance measure
  • Collision checking via safety certificates
  • For state x choose best motion primitive to get to new state y
  • Optimal path in RRT* graph
  • Cost function
• Time required for each motion primitive

Note on running additional scenarios

The given setup has been evaluated on two additional, custom dgscenaries. In order to enable them, set the "custom_cases" flag in /src/pdma4ar/exercises_def/final21/ex.py:get_final21" to True.

Progress Bar

8=============================================D
0---------------------------------------------1