From f009ed688b1de51f1f0a5bdb8095b3ac07377a09 Mon Sep 17 00:00:00 2001 From: Capri2014 Date: Mon, 28 Sep 2020 16:14:46 -0700 Subject: [PATCH] Planning & Prediction: add reference paper --- modules/planning/README.md | 14 ++++++++++---- modules/prediction/README.md | 4 ++++ 2 files changed, 14 insertions(+), 4 deletions(-) diff --git a/modules/planning/README.md b/modules/planning/README.md index ceffd35592..f5fcb908ff 100644 --- a/modules/planning/README.md +++ b/modules/planning/README.md @@ -22,7 +22,7 @@ Apollo 6.0 extended the technology to incorporate data-driven mothedologies to t ## E2E Mode ![](images/e2e_mode.png) -### How to Enable +### How to Enable - Change the configuration "learning_mode" in apollo/modules/planning/conf/planning_config.pb.txt to be "E2E_TEST" if Apollo is run in simulation or "E2E" on real vehicle - Change the configuration "model_type" in apollo/modules/planning/conf/scenario/learning_model_sample_config.pb.txt to be either "CNN_LSTM" or "CNN" and adapt the following "cpu_model_file" and "gpu_model_file" file paths. "CNN_LSTM" is the preferred model for now. @@ -48,9 +48,9 @@ The parameter `path_reference_l_weight` is for adjusting hybrid model path outpu ## Apollo 5.5 vs E2E Mode vs Hybrid Mode -We demonstrate simulation results on a dynamic nudge scenario with Apollo 5.5, E2E mode, and Hybrid mode. +We demonstrate simulation results on a dynamic nudge scenario with Apollo 5.5, E2E mode, and Hybrid mode. -- Apollo 5.5 +- Apollo 5.5 ![](images/sim_rule.gif) @@ -190,10 +190,16 @@ The Emergency scenario is another newly introduced scenario in Apollo 5.5, devel In Apollo 5.5, the Planning module architecture has been modified to incorporate new curb-to-curb driving scenarios on urban roads. As seen in the figure below, we have 2 new complex scenarios Emergency and Park-and-go. In order to plan these scenarios effectively, we have 2 new Deciders - Path Reuse Decider and Speed Bound Decider and have updated existing deciders making the planning architecture robust and flexible enough to handle many different types of urban road driving scenarios. -Each driving scenario has its set of driving parameters that are unique to that scenario making it safer, efficient, easier to customize and debug and more flexible. +Each driving scenario has its set of driving parameters that are unique to that scenario making it safer, efficient, easier to customize and debug and more flexible. > Note: > > > If you wish to include your own driving scenarios, please refer to existing scenarios as a reference. We currently do not have a template for writing your own planning scenario. ![](images/architecture_5.5.png) + + +## Related Paper + +1. [He R, Zhou J, Jiang S, Wang Y, Tao J, Song S, Hu J, Miao J, Luo Q. "TDR-OBCA: A Reliable Planner for Autonomous Driving in Free-Space Environment." *arXiv preprint arXiv:2009.11345.* ](https://arxiv.org/pdf/2009.11345.pdf) +2. [Zhou J, He R, Wang Y, Jiang S, Zhu Z, Hu J, Miao J, Luo Q. "DL-IAPS and PJSO: A Path/Speed Decoupled Trajectory Optimization and its Application in Autonomous Driving." *arXiv preprint arXiv:2009.11135.*](https://arxiv.org/pdf/2009.11135.pdf) diff --git a/modules/prediction/README.md b/modules/prediction/README.md index ffdd06e804..8f19832e67 100644 --- a/modules/prediction/README.md +++ b/modules/prediction/README.md @@ -92,3 +92,7 @@ The prediction module estimates the future motion trajectories for all perceived The prediction module also takes messages from both localization and planning as input. The structure is shown below: ![](images/architecture2.png) + +## Related Paper + +1. [Xu K, Xiao X, Miao J, Luo Q. "Data Driven Prediction Architecture for Autonomous Driving and its Application on Apollo Platform." *arXiv preprint arXiv:2006.06715.* ](https://arxiv.org/pdf/2006.06715.pdf) -- GitLab