benchmarks package

Base Classes

Benchmark

class autompc.benchmarks.Benchmark(name, system, task, data_gen_method)

Represents a Benchmark for testing AutoMPC, including the sytem, task, and a method of generating data.

abstract dynamics(x, u)

Benchmark dynamics

Parameters:

• x (np array of size self.system.obs_dim) – Current observation

• u (np array of size self.system.ctrl_dim) – Control input

Returns:

xnew – New observation.

Return type:

np array of size self.system.obs_dim

abstract gen_trajs(seed, n_trajs, traj_len=None)

Generate trajectories.

Parameters:

• seed (int) – Seed for trajectory generation

• n_trajs (int) – Number of trajectories to generate

• traj_len (int) – Length of trajectories to generate. Default varies by benchmark.

Returns:

Benchmark training set

Return type:

List of Trajectory

abstract static data_gen_methods(self)

List available data generation methods

Return type:

List of strings

Available Benchmarks

CartpoleSwingupBenchmark

class autompc.benchmarks.CartpoleSwingupBenchmark(data_gen_method='uniform_random')

This benchmark uses the cartpole system and is consistent with the experiments in the ICRA 2021 paper. The task is to move the pole from the down position to the up position. The performance metric returns 1 for every observation which is more than 0.2 away from the goal in either the angle or angular velocity dimensions, and 0 otherwise.

visualize(fig, ax, traj, margin=5.0)

Visualize the cartpole trajectory.

Parameters:

• fig (matplotlib.figure.Figure) – Figure to generate visualization in.

• ax (matplotlib.axes.Axes) – Axes to create visualization in.

• traj (Trajectory) – Trajectory to visualize

• margin (float) – Shift the viewing window by this amount when the cartpole reaches the edge of the screen

CartpoleSwingupV2Benchmark

class autompc.benchmarks.CartpoleSwingupV2Benchmark(data_gen_method='uniform_random')

This benchmark uses the cartpole system and differs from CartpoleSwingupBenchmark in that the performance metric requires the cartpole to stay within the [-10, 10] range.

visualize(fig, ax, traj, margin=5.0)

Visualize the cartpole trajectory.

Parameters:

• fig (matplotlib.figure.Figure) – Figure to generate visualization in.

• ax (matplotlib.axes.Axes) – Axes to create visualization in.

• traj (Trajectory) – Trajectory to visualize

• margin (float) – Shift the viewing window by this amount when the cartpole reaches the edge of the screen

HalfcheetahBenchmark

class autompc.benchmarks.halfcheetah.HalfcheetahBenchmark(data_gen_method='uniform_random')

This benchmark uses the OpenAI gym halfcheetah benchmark and is consistent with the experiments in the ICRA 2021 paper. The benchmark reuqires OpenAI gym and mujoco_py to be installed. The performance metric is \(200-R\) where \(R\) is the gym reward.

visualize(traj, repeat)

Visualize the half-cheetah trajectory using Gym functions.

Parameters:

• traj (Trajectory) – Trajectory to visualize

• repeat (int) – Number of times to repeat trajectory in visualization