# Myriad: a real-world testbed to bridge trajectory optimization and deep learning

Source: [Myriad: a real-world testbed to bridge trajectory optimization and deep learning](<https://www.sota2.com/research/paper/68fa6e771e209131cea40089>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-01-01T16:06:35.804Z

Authors: Nikolaus H. R. Howe, Simon Dufort-Labb\\'e, Nitarshan Rajkumar, Pierre-Luc Bacon

Published: 2022-02-22T00:47:14.000Z

## Abstract

We present Myriad, a testbed written in JAX for learning and planning in real-world continuous environments. The primary contributions of Myriad are threefold. First, Myriad provides machine learning practitioners access to trajectory optimization techniques for application within a typical automatic differentiation workflow. Second, Myriad presents many real-world optimal control problems, ranging from biology to medicine to engineering, for use by the machine learning community. Formulated in continuous space and time, these environments retain some of the complexity of real-world systems often abstracted away by standard benchmarks. As such, Myriad strives to serve as a stepping stone towards application of modern machine learning techniques for impactful real-world tasks. Finally, we use the Myriad repository to showcase a novel approach for learning and control tasks. Trained in a fully end-to-end fashion, our model leverages an implicit planning module over neural ordinary differential equations, enabling simultaneous learning and planning with complex environment dynamics.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [End-to-end learning and planning on Bacteria](<https://www.sota2.com/research/sota/end-to-end-learning-and-planning-on-bacteria.md>): End-to-end learning and planning; Bacteria; Cost (Best): -7.98.

- [End-to-end learning and planning on Bear Populations](<https://www.sota2.com/research/sota/end-to-end-learning-and-planning-on-bear-populations.md>): End-to-end learning and planning; Bear Populations; Best Cost: 12.28.

- [End-to-end learning and planning on Bioreactor](<https://www.sota2.com/research/sota/end-to-end-learning-and-planning-on-bioreactor.md>): End-to-end learning and planning; Bioreactor; Best Cost: -1.39.

- [End-to-end learning and planning on Cancer Treatment](<https://www.sota2.com/research/sota/end-to-end-learning-and-planning-on-cancer-treatment.md>): End-to-end learning and planning; Cancer Treatment; Best Cost: 20.57.

- [End-to-end learning and planning on Cart-Pole Swing-Up](<https://www.sota2.com/research/sota/end-to-end-learning-and-planning-on-cart-pole-swing-up.md>): End-to-end learning and planning; Cart-Pole Swing-Up; Cost (Best): 87.78.

- [End-to-end learning and planning on Glucose](<https://www.sota2.com/research/sota/end-to-end-learning-and-planning-on-glucose.md>): End-to-end learning and planning; Glucose; Best Cost: 1354.02.

- [End-to-end learning and planning on HIV Treatment](<https://www.sota2.com/research/sota/end-to-end-learning-and-planning-on-hiv-treatment.md>): End-to-end learning and planning; HIV Treatment; Best Cost: -823.13.

- [End-to-end learning and planning on Mould Fungicide](<https://www.sota2.com/research/sota/end-to-end-learning-and-planning-on-mould-fungicide.md>): End-to-end learning and planning; Mould Fungicide; Best Cost: 23.5.

- [End-to-end learning and planning on Mountain Car](<https://www.sota2.com/research/sota/end-to-end-learning-and-planning-on-mountain-car.md>): End-to-end learning and planning; Mountain Car; Cost (Best): 8.57.

- [End-to-end learning and planning on Pendulum](<https://www.sota2.com/research/sota/end-to-end-learning-and-planning-on-pendulum.md>): End-to-end learning and planning; Pendulum; Best Cost: 25.53.

- [End-to-end learning and planning on Predator Prey](<https://www.sota2.com/research/sota/end-to-end-learning-and-planning-on-predator-prey.md>): End-to-end learning and planning; Predator Prey; Cost (Best): 1.79.

- [End-to-end learning and planning on Timber Harvest](<https://www.sota2.com/research/sota/end-to-end-learning-and-planning-on-timber-harvest.md>): End-to-end learning and planning; Timber Harvest; Best Cost: -5104.67.

- [End-to-end learning and planning on Tumour](<https://www.sota2.com/research/sota/end-to-end-learning-and-planning-on-tumour.md>): End-to-end learning and planning; Tumour; Cost (Best): 7571.67.

- [End-to-end learning and planning on Van Der Pol](<https://www.sota2.com/research/sota/end-to-end-learning-and-planning-on-van-der-pol.md>): End-to-end learning and planning; Van Der Pol; Best Cost: 2.87.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2202.10600>)

[PDF](<https://arxiv.org/pdf/2202.10600.pdf>)

[Implementation](<https://github.com/nikihowe/myriad>)