From 970b7cb1692c9bdf12c420de96d122b8c4f0303c Mon Sep 17 00:00:00 2001 From: cgaspard3333 Date: Tue, 17 Dec 2024 11:23:39 +0100 Subject: [PATCH 1/3] Add FootstepNet Envs to doc projects page --- docs/misc/projects.rst | 14 ++++++++++++++ 1 file changed, 14 insertions(+) diff --git a/docs/misc/projects.rst b/docs/misc/projects.rst index 5f0c69710..958bcf217 100644 --- a/docs/misc/projects.rst +++ b/docs/misc/projects.rst @@ -250,3 +250,17 @@ It enables solving environments involving partial observability or locomotion (e | Authors: Corentin Léger, Gautier Hamon, Eleni Nisioti, Xavier Hinaut, Clément Moulin-Frier | Github: https://github.com/corentinlger/ER-MRL | Paper: https://arxiv.org/abs/2312.06695 + +FootStepNet Envs +---------------- + +Footsteps Planning RL Environments for Fast On-line Bipedal Footstep Planning and Forecasting. + +This project introduces an efficient and lightweight method for bipedal footstep planning in local environments containing obstacles. Leveraging state-of-the-art Deep Reinforcement Learning (DRL) techniques, our approach achieves real-time on-line inference with minimal computational requirements. Unlike traditional methods, our solution is heuristic-free and operates within a continuous action space to generate feasible and effective footsteps for navigating complex environments. + +In addition to planning, we propose a forecasting method, allowing to quickly estimate the number of footsteps required to reach different candidates of local targets. This forecasting is seamlessly integrated into the computations performed by the actor-critic DRL architecture, ensuring fast and reliable predictions without additional overhead. + + +| Authors: Clément Gaspard, Grégoire Passault, Mélodie Daniel, Olivier Ly +| Github: https://github.com/Rhoban/footstepnet_envs +| Paper: https://arxiv.org/abs/2403.12589 \ No newline at end of file From 6cae9f9a802a23b3b04f29c582e85e8de3645fb9 Mon Sep 17 00:00:00 2001 From: cgaspard3333 Date: Tue, 17 Dec 2024 11:36:55 +0100 Subject: [PATCH 2/3] Update Changelog + typo --- docs/misc/changelog.rst | 1 + docs/misc/projects.rst | 2 +- 2 files changed, 2 insertions(+), 1 deletion(-) diff --git a/docs/misc/changelog.rst b/docs/misc/changelog.rst index 1ccfbb5a1..8e7177559 100644 --- a/docs/misc/changelog.rst +++ b/docs/misc/changelog.rst @@ -39,6 +39,7 @@ Documentation: - Added Decisions and Dragons to resources. (@jmacglashan) - Updated PyBullet example, now compatible with Gymnasium - Added link to policies for ``policy_kwargs`` parameter (@kplers) +- Add FootstepNet Envs to the project page (@cgaspard3333) Release 2.4.0 (2024-11-18) -------------------------- diff --git a/docs/misc/projects.rst b/docs/misc/projects.rst index 958bcf217..62ce4cf73 100644 --- a/docs/misc/projects.rst +++ b/docs/misc/projects.rst @@ -251,7 +251,7 @@ It enables solving environments involving partial observability or locomotion (e | Github: https://github.com/corentinlger/ER-MRL | Paper: https://arxiv.org/abs/2312.06695 -FootStepNet Envs +FootstepNet Envs ---------------- Footsteps Planning RL Environments for Fast On-line Bipedal Footstep Planning and Forecasting. From 90e6c4e1e2eb52e05596e3633bfd887dfce2a645 Mon Sep 17 00:00:00 2001 From: cgaspard3333 Date: Tue, 17 Dec 2024 16:15:26 +0100 Subject: [PATCH 3/3] Update FootstepNet description --- docs/misc/projects.rst | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) diff --git a/docs/misc/projects.rst b/docs/misc/projects.rst index 4031c2dc2..48b9dc435 100644 --- a/docs/misc/projects.rst +++ b/docs/misc/projects.rst @@ -254,12 +254,9 @@ It enables solving environments involving partial observability or locomotion (e FootstepNet Envs ---------------- -Footsteps Planning RL Environments for Fast On-line Bipedal Footstep Planning and Forecasting. - -This project introduces an efficient and lightweight method for bipedal footstep planning in local environments containing obstacles. Leveraging state-of-the-art Deep Reinforcement Learning (DRL) techniques, our approach achieves real-time on-line inference with minimal computational requirements. Unlike traditional methods, our solution is heuristic-free and operates within a continuous action space to generate feasible and effective footsteps for navigating complex environments. - -In addition to planning, we propose a forecasting method, allowing to quickly estimate the number of footsteps required to reach different candidates of local targets. This forecasting is seamlessly integrated into the computations performed by the actor-critic DRL architecture, ensuring fast and reliable predictions without additional overhead. +These environments are dedicated to train efficient agents that can plan and forecast bipedal robot footsteps in order to go to a target location possibly avoiding obstacles. They are designed to be used with Reinforcement Learning (RL) algorithms. +Real world experiments were conducted during RoboCup competitions on the Sigmaban robot, a small-sized humanoid designed by the *Rhoban Team*. | Authors: Clément Gaspard, Grégoire Passault, Mélodie Daniel, Olivier Ly | Github: https://github.com/Rhoban/footstepnet_envs